AutoSync PDF

IMAGINE AutoSync
Users Guide December 2010
Copyright 2010 ERDAS, Inc.

All rights reserved. Printed in the United States of America. The information contained in this document is the exclusive property of ERDAS, Inc. This work is protected under United States copyright law and other international copyright treaties and conventions. No part of this work may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying and recording, or by any information storage or retrieval system, except as expressly permitted in writing by ERDAS, Inc. All requests should be sent to the attention of: Manager, Technical Documentation ERDAS, Inc. 5051 Peachtree Corners Circle Suite 100 Norcross, GA 30092-2500 USA. The information contained in this document is subject to change without notice. Government Reserved Rights. MrSID technology incorporated in the Software was developed in part through a project at the Los Alamos National Laboratory, funded by the U.S. Government, managed under contract by the University of California (University), and is under exclusive commercial license to LizardTech, Inc. It is used under license from LizardTech. MrSID is protected by U.S. Patent No. 5,710,835. Foreign patents pending. The U.S. Government and the University have reserved rights in MrSID technology, including without limitation: (a) The U.S. Government has a nonexclusive, nontransferable, irrevocable, paid-up license to practice or have practiced throughout the world, for or on behalf of the United States, inventions covered by U.S. Patent No. 5,710,835 and has other rights under 35 U.S.C. 200-212 and applicable implementing regulations; (b) If LizardTech's rights in the MrSID Technology terminate during the term of this Agreement, you may continue to use the Software. Any provisions of this license which could reasonably be deemed to do so would then protect the University and/or the U.S. Government; and (c) The University has no obligation to furnish any know-how, technical assistance, or technical data to users of MrSID software and makes no warranty or representation as to the validity of U.S. Patent 5,710,835 nor that the MrSID Software will not infringe any patent or other proprietary right. For further information about these provisions, contact LizardTech, 1008 Western Ave., Suite 200, Seattle, WA 98104. ERDAS, ERDAS IMAGINE, Stereo Analyst, IMAGINE Essentials, IMAGINE Advantage, IMAGINE, Professional, IMAGINE VirtualGIS, Mapcomposer, Viewfinder, and Imagizer are registered trademarks of ERDAS, Inc. Other companies and products mentioned herein are trademarks or registered trademarks of their respective owners.
Table of Contents
Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iii List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
About This Manual . . . . . . . . . . Example Data . . . . . . . . . . . . . . Documentation . . . . . . . . . . . . . Conventions Used in This Book More Information/Help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii vii vii vii viii
Introduction to IMAGINE AutoSync . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Data Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Input Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Reference Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Selecting Input and Reference Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Digital Elevation Model (DEM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 APM Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 APM Strategy Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Ideal Situations for Good APM Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Situations to Avoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 APM Trouble Shooting and Tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Image to Image (2D) Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Ground to Image (3D) Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Approximate Sensor Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Selecting a Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Selecting a DEM/DTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Modeling Trouble-Shooting and Tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 IMAGINE AutoSync Tips and Hints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Interpreting Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Using the IMAGINE AutoSync Workstation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Using the IMAGINE AutoSync Wizards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Table of Contents Table of Contents
iii iii
IMAGINE AutoSync Workflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 General IMAGINE AutoSync Tips and Hints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 General Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Using IMAGINE AutoSync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Using the Edge Matching Wizard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Using the Input tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Using the APM Strategy tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Using the Edge Match Strategy tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Using the Projection tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Using the Output tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Display Output Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 View Summary Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Using the AutoSync Workstation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Create New IMAGINE AutoSync Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Add Input Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Add Reference Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Collect Manual Tie Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Run APM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Preview Output Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Improve Output Image Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Review Image Map Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45 Set Output Image Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Resample Output Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .48 Verify Output Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .49 View Summary Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
iv
Table of Contents
List of Tables
Table Table Table Table 1: 2: 3: 4: APM Parameter Tuning . . . . . . . Tie Point-Based Model Selection Edge Match Image Data Set . . . Georeference Image Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 17 23 33
List of Tables List of Tables
v v
vi
List of Tables
Preface
About This Manual
The IMAGINE AutoSync Users Guide serves as a handy guide to help you use IMAGINE AutoSync. Included is a comprehensive index, so that you can reference particular information later.
Example Data
Sample data sets are provided with the software. This data is separately installed from the data DVD. For the purposes of documentation, <ERDAS_Data_Home> represents the name of the directory where sample data is installed. The Tour Guides refer to specific data which are stored in <ERDAS_Data_Home>/examples.
Documentation
This manual is part of a suite of on-line documentation that you receive with ERDAS IMAGINE software. There are two basic types of documents, digital hardcopy documents which are delivered as PDF files suitable for printing or on-line viewing, and On-Line Help Documentation, delivered as HTML files. The PDF documents are found in <IMAGINE_HOME>\help\hardcopy where <IMAGINE_HOME> represents the name of the directory where ERDAS IMAGINE is installed. Many of these documents are available from the ERDAS Start menu. The on-line help system is accessed by clicking on the Help button in a dialog or by selecting an item from a Help menu.
Conventions Used in This Book
In ERDAS IMAGINE, the names of menus, menu options, buttons, and other components of the interface are shown in bold type. For example: In the Select Layer To Add dialog, select the Fit to Frame option. When asked to use the mouse, you are directed to click, Shift-click, middle-click, right-click, hold, drag, and so forth. clickdesignates clicking with the left mouse button. Shift-clickdesignates holding the Shift key down on your keyboard and simultaneously clicking with the left mouse button. middle-clickdesignates clicking with the middle mouse button. right-clickdesignates clicking with the right mouse button. holddesignates holding down the left (or right, as noted) mouse button.
Preface Preface
vii vii
dragdesignates dragging the mouse while holding down the left mouse button.
The following paragraphs are used throughout the ERDAS IMAGINE documentation: These paragraphs contain strong warnings.
These paragraphs provide software-specific information.
These paragraphs contain important tips.
These paragraphs lead you to other areas of this book or other ERDAS manuals for additional information. NOTE: Notes give additional instruction. Shaded Boxes Shaded boxes contain supplemental information that is not required to execute the steps of a tour guide, but is noteworthy. Generally, this is technical information.
More Information/Help
On-Line Help
As you go through the tour guides, there are several ways to obtain more information regarding dialogs, tools, or menus, as described below.
There are several ways to open Help:
Click the Help icon Click the Help button on any dialog Read the Bubble Help
viii
Preface
Help Icon Clicking this icon launches On-Line Help at the main start page. This file provides access to all On-Line Help through the Contents, Index, and Search tools. From the Contents pane, you may jump to any of the available On-Line Help documents. From the Search pane, you may search the entire On-Line Help system. Help Button Dialogs within ERDAS software contain a Help button which you can click to open On-Line Help for the respective dialog. To open the navigation pane (containing the Contents, Index, and Search functions) you must click the View Navigation icon Bubble Help As you move your cursor moves along individual tools on many of the tabs, a small Bubble Help popup dialog may open, if it is available for the item and if your Preference is set to use Bubble Help. This information is the same as the Status Bar Help shown in the bottom of the Workspace, but opens directly under your cursor. at the top of the help page.
Preface
ix
Preface
Introduction to IMAGINE AutoSync

This chapter explains what to expect from IMAGINE AutoSync and how to adjust the underlying engine for optimal results. It also provides practical tips and hints, and describes the best strategies to handle difficult situations you may encounter.
Overview
Imagery always needs to be geometrically corrected to a map coordinate system to be useful. This is especially true for applications such as change detection, resolution merge (pan sharpening), mosaic, and even simple layer stacking. The geometric correction must be highly accurate, because misalignment of features at the same location could render the results useless. Additionally, many applications require the creation of a large database of georeferenced images. The current process of manual point measurement can be prohibitively labor intensive for large applications, and it does not enforce sub-pixel level correlation between images due to the limitation of human visual interpretation. Block triangulation, although tying imagery together photogrametrically, does not enforce any correlation to already existing image layers. IMAGINE AutoSync uses an automatic point matching algorithm to generate thousands of tie points, and produces a mathematical model to tie the images together. The resulting workflows significantly reduce or sometimes completely eliminate manual point collection. The output should be generally equal or better in accuracy in comparison to the current methodology.
Benefits
Some of the benefits of using IMAGINE AutoSync include: Automatic point matching algorithm for mass tie point generation. Automatic sensor detection (if sensor information is available). Streamlined wizard workflow that can handle a large number of similar images. Powerful workstation that provides an efficient environment and tools for tie point quality assessment, point measurement, preview of the output, and is useful in the organization of complex projects, and so on. Iterative model refinement and instantaneous rectification results in the workstation environment.
Constraints
As with any tool, poor data quality and/or inappropriate parameters can produce less than desirable results. As a user, you should be aware of the following:
Introduction to IMAGINE AutoSync Introduction to IMAGINE AutoSync
1 1
Default parameters are appropriate for most data, but may require adjustment to work well for more varied scenes. Data quality of the input images, reference image, and Digital Elevation Model (DEM) directly affects the final output results. User intervention may be required in some workflows to compensate for poor data quality, erroneous model selection, incorrect tie point measuring, and so on.
The more experienced you are with using IMAGINE AutoSync and the more knowledge you have about the data and workflow, the better the output will be using our software.
Data Preparation
The quality of your input data plays a crucial role in determining the accuracy of the output and extent of user intervention required. Additionally, the type of the input data largely determines which workflow to follow for optimal results. This section discusses various data you can use in IMAGINE AutoSync and how to best prepare the data. It also provides suggested remedies for potential problems. When using the edge matching workflow, you can use georeferenced or calibrated input images. In the georeferencing workflow, input images can be georeferenced, calibrated, or raw images. You can also use images that have map information but are not georeferenced to a particular projected coordinate system. If you are using raw input images, you must first establish a footprint with the reference image before running automatic point measurement. This is a necessary step since raw images lack the map information to place the image at an approximate location to overlap with the reference image. Another consideration when using raw imagery is the potential for matching problems between the uncorrected, vertically displaced mountainous regions in the raw image and an orthorectified reference image. This displacement can cause poor points to be generated from the automatic point measurement process. You can alleviate this problem by choosing an appropriate sensor model that allows for the specification of a DEM (DLT, RPC, or ROP) and using an accurate DEM. See Modeling on page 13 for more details. Sensor metadata can be very helpful in establishing models for rectification. For example, QuickBird images contain enough information to build a rigorous model. Generally, data that are rectified using a rigorous model and an accurate DEM produce the best results.
Input Images
Reference Images
The reference image must be georeferenced and have projection information. For best results, the reference image should have a higher or spatial resolution equal to the input images. IMAGINE AutoSync has been tested with datasets that do not adhere to the above suggestion and at a factor of six difference, it was generally found that tie point editing was required to improve the solution. When selecting input and reference images to use, it is preferable to use images with maximum similarity. This largely improves the result of the automatic point measurement. If the images are too dissimilar, the results from the automatic point measurement process may be undesirable. Some of the main factors that affect the similarities of images include the following:
Time of Capture
Selecting Input and Reference Images
Time of capture, especially the season, could greatly alter the radiometric characteristics of the images. For example, a winter scene will not match well with a summer scene with high vegetation.
Resolution
Resolution is another factor that affects point matching results, because it creates a difference in the details of the two images. Avoid mixing input and reference images with a resolution difference larger than a factor of six.
Elevation Variation
Variation in the elevation could also cause a difference between the input and reference images. This is because the reference most likely will be an orthorectified image and therefore vertical displacement is minimal compared to the input. As a result, features that should be in the same location could be far apart when the input and reference images are attempted to be matched. To avoid this problem, you can select a model that will allow for the specification of a DEM.
Spectral Range of Selected Band
The automatic point matching process uses a single band of a multiband image to locate the tie point pairs. Consequently, it is important to make sure similar bands are selected from the input image and the reference image to achieve maximum similarity in radiometric characteristics. Selecting a band within the visible spectrum will produce the best results. Infrared and thermal bands should generally be avoided.
Sensor
The sensor used in capturing the image (Landsat, IKONOS, SPOT, and so on) affects the resolution and radiometric characteristics of the images. It also determines the mathematical models that can be used in the rectification. For example, the Rigorous Orbital Pushbroom sensor model can only be applied to sensors such as IRS, QuickBird, and so on. See Modeling on page 13 for details on selecting an appropriate model.
Digital Elevation Model (DEM)
The availability of a high resolution Digital Elevation Model (DEM) can drastically impact the quality of rectification results, especially for mountainous areas. A DEM provides additional model-solving information in determining the location of features in the output. This could greatly reduce the negative impact of vertical displacement when matching input and reference images.
APM Engine
Automatic Point Measurement (APM) is a software tool that uses image-matching technology to automatically recognize and measure the corresponding image points between two raster images. In IMAGINE AutoSync, APM aims to deliver the coordinates of evenly distributed corresponding points between an input image and a reference image. APM works automatically to find the needed image points, but there are a set of parameters you can adjust in circumstances where the default settings fail to produce acceptable results. An Advanced Point Matching Strategy dialog is also provided for more control over the process.
APM Strategy Parameters
The defaults that appear in these dialogs can be set from the IMAGINE AutoSync category of the Preference Editor.
Before using APM, confirm that the Initial Pyramid Layer Number is set to 1 in Image Files (General) category of the Preference Editor. This retains the largest pyramid layer when computing pyramid layers to ensure point matching accuracy.
APM Strategy Tab
You can adjust APM parameters on the APM Strategy tab on the IMAGINE AutoSync Project Properties dialog in the workstation, or on the APM Strategy tab in the wizards.
Input Layer to Use: Select the layer of the input image to use for APM. IMAGINE AutoSync automatically assigns a layer name for each layer in the input image, using the following format: Layer_1, Layer_2, and so on. Layer 1 refers to band one of the image. If you have multiple input images in your list and they contain a different number of bands, all possible bands will be listed. If you choose Layer_5 and one or more images only contain four bands, band 5 will be used on any images containing 5 bands, but band 4 or the next available band will be used on any images with fewer than 5 bands. Reference Layer to Use: Select the layer of reference image to use for APM. IMAGINE AutoSync uses the actual reference layer names.
Try to select the layers of two images with same or similar spectrum. For example, if you use two color images, then try to choose the same green bands. The more similar the radiometric characteristics of two images are, the better APM results it can achieve. In case you are using images with different band widths (such as a black-white image and a multi-spectral image), you may still get good results, but you may also encounter less accurate matching points depending on the specific situation. Find Points With: Select either the Default Distribution or Defined Pattern type of point distribution. Default Distribution APM superimposes a regularly spaced grid onto each image in an attempt to find matching points in a well distributed pattern. When you are attempting to collect fewer than 100 points, a 5 x 5 grid is used. If you want to collect 100 points or more, a 10 x 10 grid is used.
APM tries to collect matching points within an area of 512 X 512 pixels centered on the corresponding grid intersection of each image. If a matching point is not found within this area, APM moves to the next grid intersection. This optimizes the likelihood of finding well distributed points while minimizing search time on large images.
10% 30% 50% 70% 90% 10% 30% 50% 70% 90%
The 5 X 5 grid divides each image as shown at left. The 10 X 10 grid begins at 5% and increments by 10%. In both cases, the maximum search area is 512 X 512 pixels centered on each grid intersection.
Default distribution is most appropriate for standard aerial photography and less so for the more varied possible inputs to IMAGINE AutoSync. For that reason, Default Distribution is available as an option to those working with aerials, but the standard default setting for finding points is set to Defined Pattern in anticipation of a wider use of data. Defined Pattern If you select Defined Pattern, the options Starting Column, Starting Line, Column Increment, Line Increment, Ending Column, and Ending Line become active. Using those options, you can define the exact placement of tie points throughout the images of the block. With this option, you can also define the intended number of points per pattern. Intended Number of Points/Pattern (Image) This section of the APM Strategy tab changes depending on whether you have selected Default Distribution or Defined Pattern, above. In the case of Default Distribution, enter the intended number of tie points generated for each image. The minimum is 9 and the maximum is 500. The default is 400. In the case of Defined Pattern, enter the intended number of tie points generated per pattern. The minimum number of points per pattern is 1 and the maximum is 8. The default is 1.
If you want to define a tie point pattern, select Defined Pattern. However, you should consider the overlap percentage when you define your own pattern and try to get each pattern location inside as much of the overlap as possible. Keep All Points Select this checkbox to use all tie points generated regardless of accuracy or distribution. If this checkbox is active, the number of collected tie points will be greater than the intended number of points per image. You do not normally need to choose this option unless your images have low contrast, yielding few points without this option selected. Starting Column, Starting Line: Define the starting location of tie points you want to find on the image in pixels. You will get better results if you define the starting location close to the upper-left corner of the overlap area on the higher resolution image. It is safe to define the location close to the upper-left corner of the image, but you may get bad results if it is close to the lower-right corner. Column Increment, Line Increment: Define the increment in pixels for tie point locations along column and line direction. APM will try to find tie points around the image locations with these increments to the previous locations. Ending Column, Ending Line: Define the last column and line for tie point collection. If you want to define them, they should not exceed the lower-right corner of the overlap area. If you leave them at the default of 0, 0, APM will automatically use the last column and last line of the overlap area. Automatically Remove Blunders: Click this checkbox to remove blunders (wrong tie points) automatically from the APM generated tie points. Removing blunders is an iterative process based on a 3rd order polynomial model. When this option is selected, the points that do not fit well with the majority of tie points are considered blunders and are discarded. By default, this option is selected. You should deselect this option only if you suspect that it is removing correct tie points. For example, you should deselect this option when most of the APM tie points are wrong, or when there is a large difference in the terrain between the inputs and reference image. Maximum Blunder Removal Iterations: This option becomes available when you choose to automatically remove blunders with the Automatically Remove Blunders option. The default is 2. In most cases, increasing this number means more iterations of the blunder removal algorithm will be run. As a result, more tie points will be considered as blunders and discarded.
Advanced Point Matching Strategy Dialog
On the Advanced Point Matching Strategy dialog in IMAGINE AutoSync, you can adjust the more advanced APM parameters to optimize automatic tie point collection.
Search Size: Enter the window size in pixels to use for searching for corresponding points. IMAGINE AutoSync searches for the corresponding point within a square window defined by this parameter. The default value is 17 (a 17 x 17 pixel window). For flat areas, this value could be smaller, for steeper areas, it could be larger. A larger value could cause more computation time and more wrong points, but a smaller value could result in fewer matched points. Feature Point Density: Defines the feature point density percentage based on an internal default. Increasing the value above 100% (for example, in poor contrast area) will get more feature points to get more matched points. Decreasing the value below 100% (for example, in an area with crowded details) will result in fewer feature points to accelerate the computation.
Normally, you do not need to adjust this parameter if you are using scanned aerial photos. However, if you select the Avoid Shadow option, you should set this parameter to a higher value (for example, 200%). Correlation Size: Enter the window size in pixels for crosscorrelation. The default window size is 11 (11 x 11). A larger window size could cause a smaller correlation coefficient due to the geometric difference within the two correlation windows and, therefore, fewer matched points. A smaller window size could result in a larger correlation coefficient due to insufficient contents, therefore yielding more bad points.
Minimum Point Match Quality: Enter the limit for the cross correlation coefficient. The allowed value is between 0.6 and 0.99. The default value is 0.80. A larger limit results in fewer points accepted and less error. A smaller limit results in more correlated points but possibly more errors. Least Squares Size: Enter the window size in pixels for least square matching. A larger window size could reduce the number of badly matched points, and could also reduce the number of good points. A smaller window size could increase the number of both bad and good points. The default is 21. You should increase this number for flatter areas and decrease the number for steeper areas. Setting the window size too small could result in insufficient contents in the window for the least square computation. Initial Accuracy: Enter the relative accuracy of your initial values used by the automatic tie point generation process. Generally, a large value here increases the initial search area for possible corresponding points at the initial estimation phase. This value can be seen as the relative accuracy of the source you have chosen for your initial values (for example, initial map coordinates or relative terrain change). The default value is 10%. You should use initial values with an accuracy of 10% or better for the automatic tie point collection. Avoid Shadow: When you select this checkbox, APM will try to avoid generating tie points in areas of shadow. Avoiding areas in shadow improves tie point accuracy. You should choose the type of images you are working with; negative or positive image scans.
You do not need to use this option unless shadows are very prominent in your images. Image Scan Type: Positive Select this option if you are working with a positive image (bare ground appears white in the image). Image Scan Type: Negative Select this option if you are working with a negative image (bare ground appears black in the image). Use Manual Tie Points for Initial Connection between Images: Select this option if the manually measured tie points will be used as the initials for APM to find additional tie points automatically. Select this option when the initial map coordinates are very coarse, or when no map information is available for one or more of the images. If you try to rectify a raw image to another raw image, you should also select this option. You should manually collect a minimum of three points.
Exclude Background Area: Select this option if you want to exclude the background of the image. When you select this option, a bounding box excluding the background area will be calculated and used as the active image area for APM. The default starting column and starting row will move inside of this bounding box. If you manually changed the values in the Starting Column and Starting Row, then your new values will take precedence above the calculated bounding box. Background Value: This option becomes available when you select the Exclude Background Area option. Enter the background value of the image. The default value is zero. If you do not know the background value, you can use the IMAGINE ImageInfo tool and review the pixel values of the image.
Ideal Situations for Good APM Performance
For the best APM results, try to ensure that the following conditions are met as much as possible. Not meeting one or more of these conditions does not necessarily mean that the APM results will be of poor quality. Use images with an overlap larger than 40%. Use images with the same or similar resolution or pixel size. Use images that were captured in the same season, at the same time of day (similar illumination conditions), and with similar weather situations with good visibility. Use images that were captured by the same or similar sensor. Select the same band or a similar band in the images for point matching to ensure similarity of radiometric characteristics. Use images that are properly orthorectified (if appropriate). This reduces the impact of vertical displacement and other distortions. Pay special attention to quality of orthorectification. A poorly orthorectified image produces bad results and is misleading in raising your expectation. Use images with relatively flat terrain. There is minimal vertical displacement and the radiometric characteristics are better preserved because they have not gone through extensive modification in a prior rectification process. Ensure good initial map information is available for the images. Images with less than 10% misalignment in the overlap region tend to yield better results. When there is no initial image map information, you need to perform an initial manual registration. You can do this by digitizing 3-4 high quality points that are evenly distributed and preferably placed close to the image corners.
10
When using images with mountainous terrain, use an accurate DEM in order to remove the image displacement caused by the steep terrain change. Before using APM, confirm that the Initial Pyramid Layer Number is set to 1 in Image Files (General) category of the Preference Editor. This retains the largest pyramid layer when computing pyramid layers to ensure point matching accuracy.
Situations to Avoid
These are some situations you should try to avoid, since any or a combination of the following conditions could result in poorly matched points.
If you are unable to improve any of the following conditions, refer to APM Trouble Shooting and Tips on page 12 to learn about possible remedies by adjusting the APM parameters. Using images with an overlap that is less than 256 x 256 pixels or with an overlap region that is too narrow. Since APM requires a sufficient region to deploy the matching strategy, an overlap less than 20% will not produce desirable results. Resolution differences (or pixel size difference) larger than 6 times. This is the threshold for any meaningful results. Using images with a drastic time difference can result in a poor match. For instance, a winter scene does not match well with a summer scene due to the change in vegetation. Images do not match well if they were captured too many years apart if there has been a high level of change. The band to match selection for inputs and reference should not differ too much in electromagnetic wavelength. For example, an infrared band might not match well with a blue band. Sensor characteristics between the images if too different can affect performance. For instance, a very high flying sensor does not match well with a low flying sensor. The band width used, sensor mechanism (pushbroom or discrete sensors) all could play a role in making the images distinct from each other. If using images with large, uncorrected terrain relief, the vertical displacement can drastically reduce the likelihood of successful APM results. Large scale, high-rise urban scenes do not match well due to the vertical displacement effect, which is difficult to orthorectify.
11
Bad initial image registration. If misalignment in the overlap region is larger than 20%, or manually digitized points are unevenly distributed too far from image corners or of poor quality, results may be affected. Inappropriate placement of the starting point too close to the ending edge (lower-right corner) of overlap area. This will result in the overlap area not having enough search regions.
APM Trouble Shooting and Tips
Refer to this section if the APM results are not as expected. If APM results in a large RMSE, it may indicate bad APM results and/or inappropriate modeling. Examine the tie points carefully to ensure that the problem is from the APM results (many bad points, not enough points, and so on.) before applying the following steps to fine tune the APM parameters for improved results. If the APM points are correct but the output does not reflect the quality of the points, you most likely have chosen an inappropriate model.
Problem Diagnosis and Solutions
Sometimes you may not be able to rectify or improve any of the conditions. If that occurs, you can adjust the APM parameters by following the steps listed in the next section.
APM Parameter Tuning
When you need to adjust the APM parameters, first match your situation with one of the following, and then adjust the APM parameters accordingly. Table 1: APM Parameter Tuning
Situation
Many points, but many poor quality points
Remedies
On the APM Strategy tab, change one or more of the following parameters:
Increase the Minimum Point Match Quality (> 0.9) Increase the
Correlation Size and Least Squares Size Intended Number of Points
Decrease the
12
Situation
Too few points, but good quality
Remedies
On the APM Strategy tab, change one or more of the following parameters:
Increase the Decrease the
Intended Number of Points Column Increment and Line
Increment
Too few points, and On the APM Strategy tab, change one or more of the poor quality following parameters:
Increase the Decrease the Decrease the Increase the
Search Size Correlation Size Least Squares Size Intended Number of Points
NOTE: Most likely, you have one or more undesirable conditions, such as a large misalignment. If the problem is a large misalignment, manually collect a few tie points and select the Use Manual tie points for initial connection between the images option under APM Strategy | Advanced Options.
Too many points On the APM Strategy tab, change one or more of the following parameters:
Decrease the Increase the
Intended Number of Points Column Increment and Line
Increment
Increase the Minimum Point Match Quality (> 0.9) and increase the Correlation Size and
Least Squares Size NOTE: A large of number of points is unnecessary and slows down the APM processing. A large number of points does not necessarily improve the accuracy. Try to trim down the numbers and keep the best points.
13
Edge Matching
The nature of APM is to avoid edges of images. Features on the edge, if no points were found, could be misaligned. For the edge matching workflow, it will typically align features towards the center of the overlap. If the next step of processing is mosaicking, a cutline down the center of the overlap will eliminate issues on the edge. To match the entire overlap, it may be necessary to manually collect points on the edge, especially on linear features, to obtain full alignment.
Modeling
This section provides a brief explanation of the various mathematical models available in IMAGINE AutoSync. You can apply these models to the input images in order to geometrically correct them. Understanding the model properties helps you select the model that will generate the most accurate results for your dataset. An image-to-image transform can warp one image onto another without the use of an earth model (DEM). The fit will not be as good as a rigorous sensor model that requires a DEM because much of the distortion comes from the terrain.
Rubber Sheeting
Image to Image (2D) Transforms
Rubber sheeting is a two dimensional image-to-image transformation which is implemented as a piecewise transformation based on the triangles formed from the tie points. This has the property that the transformation is always perfect at the control points and there is always a well behaved transition from triangle to triangle. However, if an image has hilly or mountainous terrain, you will have to collect a large number of tie points. In effect, the tie points will be forming a model of the terrain surface. This can be impractical since the performance of rubber sheeting decreases as the number of points increases. Rubber sheeting is best used in an area of moderate relief when an actual sensor model and a DEM is not available.
Polynomial
A polynomial model is a two dimensional image-to-image transformation. The polynomial model is of the form: If the order is set to 1, then the result is an affine or linear transformation that is appropriate for cases where there is little or no terrain displacement and most of the image to image difference is in the form of scale, offset, and rotation. You can use higher orders in cases where there is slowly varying terrain effects with scale, offset, and terrain rotation. In both cases, it is assumed that the actual sensor model and DEM are not available.
14
X = Y =
Ai x y
i
i (n i)
Bi x y
i
i (n i)
Ground to Image (3D) Transforms
Every image is a mapping of three-dimensional (3D) coordinates into a two-dimensional (2D) plane. The ground-to-image transformation models this mapping using a DEM as the earth model.
Rigorous Orbital Pushbroom (ROP)
Many current satellite imaging systems use pushbroom sensors. This is a sensor that has a linear sensing array associated with an optical system. The whole system is moved forward by the orbital motion of the satellite. As it moves, the single line is pushed forward, scanning a whole image. The linear array can be thought of as a very narrow camera, which can be modeled with the same six parameters as a frame camera. Each line has a different set of six parameters. Because the orbit of the satellite is very stable and well known, the position (x, y, z) and orientation, omega, phi and kappa (, , ) can be modeled as time varying parameters. The time is related to each line so these parameters can be computed give the line number in the image. In the case of the Rigorous Orbital Pushbroom (ROP), the parameters of the orbit model are refined using the tie points.
Approximate Sensor Models
It is possible to use mathematical approximations for many sensor models. These approximations do not directly model the sensor. Instead, they are based on mathematic formulas whose results are very close to the rigorous sensor model.
Projective Transform Model
The Projective Transform model provides a more powerful modeling capability for multiperspective satellite images such as Landsat, SPOT, and QuickBird. While the best way to transform is to use the sensorspecific models, the Projective Transform can be used in situations where no ephemeris is available, where there is no applicable sensor model, or where the satellite image has already been geometrically corrected.
15
Rational Polynomial Coefficients (RPC)
Rational Polynomial Coefficients are (as the name implies) ratios of polynomials. These can model reasonably complex transformations and remain stable (high order polynomials tend to be unstable). Like a polynomial, a rational polynomial is described by an order and a set of coefficients for the numerator and denominator for the X and Y term. A rational polynomial is a 3D to 2D transformation in the sense that a ground x, y, and z are used along with the coefficients to compute the image row and column values. The RPC coefficients are typically computed using a solved rigorous model. One reason to use RPCs is that you do not need to know the original rigorous model, so it is a good way to provide a common framework which is independent of the actual sensor used. The RPC values themselves cannot be computed from the tie points and ground points. However, existing RPC values can be refined to provide more accurate transforms.
Direct Linear Transform (DLT)
The Direct Linear Transform (DLT) is actually an RPC whose order is equal to one. The DLT is an excellent approximation for frame cameras, and when it is known that the data comes from a frame camera, you can use this without knowing the specifics of the frame camera. The DLT coefficients can be computed from the tie and 3D points.
Selecting a Model
You will get the best results when using a rigorous sensor model and an accurate DEM. Most of the satellite data are shipped with sensor model data (either parameters for the rigorous orbital pushbroom or RPCs) which IMAGINE AutoSync can read. If the model is unknown but a DEM is available, then it is a reasonable strategy to first try using a DLT. If the results from the DLT are not acceptable, the image may have been created using a pushbroom sensor. In this case, the pushbroom orbital parameters are unknown, so the next best candidate is to use one of the image-to-image (2D) transformations as described above. The quality and accuracy of the results will be directly tied to the quality of the DEM or DTM (Digital Terrain Model) used. A DTM usually does not include man-made structures such as buildings or bridges, so it can be expected that these features will have the most mismatch in the final results. Refer to this section for additional modeling troubleshooting help and tips. The more rigorous the model, the better the result. Follow the list to make the best of the available information from the data. The recommended order of models from the most rigorous to the least is:
Selecting a DEM/DTM
Modeling TroubleShooting and Tips
16
1. Rigorous Orbital Pushbroom (ROP) 2. Rational Polynomial Coefficients (RPC) 3. Direct Linear Transform (DLT) 4. Polynomial 5. Rubber Sheeting
The quality of the points is essential in determining the outcome of the image-to-image transformation models. Ideally they should be evenly distributed and closely matched. If a particular region has very few points, you may collect some points manually to compensate. If you determine that image to image (2D) transform is the best available model for the images, the following tips can help you decide whether to use an Affine, Polynomial, or Rubber Sheeting model. Table 2: Tie Point-Based Model Selection
Number of Tie Points

< 10 1050 > 50
Appropriate Model
Affine Polynomial (3rd Order) Polynomial or Rubber Sheeting Note: When the density of points is satisfactory with an even distribution, use Rubber Sheeting. Otherwise, use a 3rd Order Polynomial.
Rubber sheeting models always have an RMSE of zero. Points would need to be manually reviewed in order to assess accuracy. When using the Linear or Non-Linear Rubber Sheeting model, you can first try using a Polynomial model for the data. Then use the RMSE as threshold to find and remove any mismatched points using the RMSE threshold selection tool in the IMAGINE AutoSync Workstation. The RMSE threshold selection tool will select all points that do not meet the specified value. By right clicking in the far left column on the CellArray, you can delete all selected points quickly eliminating the bad points. Finally, apply the Rubber Sheeting Model. Only the remaining points will be used, yielding better results.
17
When results from using the Linear or Non-Linear Rubber Sheeting model show significant misalignment in a region, selectively collect some manual tie points in that area and resolve the model to correct the situation. This process can be very effective in correcting a situation where the overall model fits well, but is problematic in a small number of regions. Deleting points with an Error Mean + 2 x Standard Deviation, and resolving the model 3 to 4 times will drastically improve the overall error results. Repeating this process more than 3 to 4 times will have diminishing results. This is more applicable when using a rigorous model such as RPC, Orbital Pushbroom, or DLT. When using other models (Polynomial, Linear, or Non-Linear Rubber Sheeting, and so forth), this method may not be as beneficial. During the process of resolving the model, you can undo the last deletion of points if the model results are not what you expected.
IMAGINE AutoSync Tips and Hints

Interpreting Results
This section provides additional tips and hints for using IMAGINE AutoSync to generate the best results.
After careful data preparation, you can run APM and tie the images together through a mathematical model. Then you can review the results. This section explains how to correctly interpret the results, identify any problems, and how to resolve them.
Visual Inspection
Visual inspection in the workstation is the most reliable method to verify results. Use the Swipe tool on overlaying images to inspect them for proper alignment. A well-aligned set of images will swipe smoothly without sudden visual interruption, except where there are real changes (for example, new buildings) or shadows.
Tie Point Quality Analysis
Inspect selected areas and use the Zoom tool in the workstation to assess the quality. Also look for an uneven distribution of tie points. If this occurs, you may need to manually collect some points to compensate. An easy way to locate suspect points is to select points based on RMSE (for instance, RMSE > Error Mean + 2 x Standard Deviation). Then use the Drive To tool to locate the points quickly.
18
RMSE Analysis
RMSE is the cumulative result of point matching and modeling. A large RMSE could be caused by one or both. Inspect the tie points to determine whether the points are the culprit. Often times, the inappropriate use of a model is responsible for a large error value. An example would be when the point quality is very good, yet the results show a large RMSE. Conversely, a small RMSE may not necessarily indicate good overall results. It could be an artifact of the model that is being used. For example, the Linear or Non-Linear Rubber Sheeting model by definition will produce a zero error because all tie points are meant to match exactly in the results while the regions surrounding the points may be distorted. Therefore, you should always perform a visual inspection along with analyzing the RMSE to ensure correct judgment of the error conditions.
Follow-up Actions
When your analysis of the results point to problems either in APM or modeling, refer to the proper sections of this chapter for specific tips for improvement: For better data preparation, refer to Data Preparation on page 2. For APM parameter tuning, refer to APM Engine on page 4. For a better model choice, refer to Modeling on page 13.
19
Using the IMAGINE AutoSync Workstation
This section provides some helpful tips when using the IMAGINE AutoSync workstation. Use the IMAGINE AutoSync Workstation for complex workflows that require more user intervention. The workstation provides more flexibility, tools for visual inspection of the results, and for manually collecting tie points. Set the color of the GCPs in the IMAGINE AutoSync preferences before running APM in the workstation. You will avoid having to select all of the GCPs individually generated by APM to change their color. Selecting a large number of GCPs (> 2000) can be very timeconsuming. Use the Preview Output option on the Input Image context menu to view the results of the model before calibrating or resampling the imagery. While in the preview mode, you can continue to delete GCPs and resolve the model. Also, whenever you select Preview Output again, the model will be recomputed and the viewer updated appropriately. This avoids having to return to the Point Review mode. When both the input and reference image contain projection information and the two images are of different resolutions, click the Set Same Scale icon on the toolbar in the workstation to set the display scale of each image to be the same. During point review, it will be easier to find similar features between the two images. If large errors are produced when using imagery with no projection information, review the manually collected points to ensure they were collected over the same features. Low quality manual GCPs will produce low quality models. Using very low resolution elevation data with the rigorous models may result in a shearing affect in the output imagery. This is due to the difference in the resolution of the input image and the elevation data. It saves time to turn off the display of a large number of tie points in the Overview in the workstation.
Using the IMAGINE AutoSync Wizards
This section provides some helpful tips when using the IMAGINE AutoSync wizards. Use the Georeferencing or Edge Matching wizards when you want to be guided through a workflow.
20
The wizard workflow can be labor-saving when the datasets are homogenous, large, and can be processed with the same settings of parameters and models. Do not mix different types of datasets where each dataset requires different settings in the same wizard workflow. Separate the datasets into different projects. Before batching large homogenous datasets with the wizard, try to run one dataset through the wizard and inspect the results in the workstation. Experiment with the settings of the workflow to find one that generates optimal output. Then batch the large datasets with the same settings. After you determine a good workflow, you can make batching easier by creating a template IMAGINE AutoSync .lap file with the proper settings but no images in the workstation. Then load the template .lap file in the wizard and add the large datasets.
IMAGINE AutoSync Workflows
IMAGINE AutoSync supports three main types of workflows. Use the workflow that is suitable to the nature of the data and your applications.
Georeferencing Workflow
Use the georeferencing workflow if you know that one input image is clearly of better accuracy, and both images are georeferenced. For example, use the georeferencing workflow if you have a database of high-accuracy images and you want to introduce another georeferenced image of lesser quality to the database.
Edge Matching Workflow
Use the edge matching workflow to bring the overlap area of image pairs into alignment. This workflow will modify both images due to the fact that the required shifts will be divided between the images. Edge matching may also be a good choice when the overlap area is small and it is undesirable to apply the same transform that is suitable for the overlap region to the entire image.
Raw Imagery Workflow
Use the raw images workflow when an image does not have georeferencing information available, or when it is unreliable. When the georeferencing information is unreliable, you should ignore the existing georeferencing information. With raw images, you need to manually collect at least three tie points that are evenly distributed, close to the image corners, and high quality.
General IMAGINE AutoSync Tips and Hints
Some general tips and hints for using IMAGINE AutoSync include:
21
If you calculate statistics on all imagery before using the image within IMAGINE AutoSync, this will improve the overall performance. IMAGINE AutoSync will recompute pyramids at the start of the APM process for any image that does not have 3x3 pyramids previously computed. ERDAS IMAGINE defaults to producing 2x2 pyramid layers. You can change the preference in Image Files (General) category of the Preference Editor to always produce 3x3 to save time. Before using APM, confirm that the Initial Pyramid Layer Number is set to 1 in Image Files (General) category of the Preference Editor. This retains the largest pyramid layer when computing pyramid layers to ensure point matching accuracy. Work with local files whenever possible. Saving IMAGINE AutoSync project files that contain a large number of GCPs may be slow over a network. When resampling, ensure the output cell size is reasonable for the images. The IMAGINE AutoSync defaults may not be suitable for your application.
Summary
When properly used, IMAGINE AutoSync is a powerful tool for fast image rectification with a tremendous saving of manual labor. This is achieved by a streamlined workflow, user-friendly workstation environment, a state-of-the-art automatic point matching engine, and a wide selection of intelligent modeling methods. The final output from IMAGINE AutoSync is the cumulative result of the workflow you select, the data quality, APM engine usage (parameter settings), and the model selected. To ensure the best results, you should make careful and judicious decisions on these factors, starting with the data preparation, and proceeding with the steps as outlined in the sections of this chapter. As with any sophisticated system, using IMAGINE AutoSync requires that you have a basic understanding of the various components of the embedded technologies. The more knowledge you have with regard to the data and the internal working of IMAGINE AutoSync, the better your chance of success.
General Guidelines
Some general guidelines you should follow when using IMAGINE AutoSync include:
1. Start with careful data preparation to ensure that you obtain the best data available.
22
Refer to Data Preparation on page 2.

2. Ensure that you understand the parameters of the APM engine. Analyze the data to see how many ideal or undesirable scenarios the data may exhibit and try to rectify them.
Refer to APM Engine on page 4.

3. Select the most accurate model for rectification and utilize the provided metadata. Understand the limitations of each model and troubleshoot accordingly
Refer to Modeling on page 13.

4. Follow the tips and hints. This will help you avoid frustration caused by improper use.
Refer to IMAGINE AutoSync Tips and Hints on page 17.

5. If your APM results are not as expected, analyze whether it is the result of bad tie points or improper choice of a sensor model. Then proceed to rectify the situation accordingly.
Refer to Modeling on page 13.
23
Using IMAGINE AutoSync

Introduction
IMAGINE AutoSync provides both wizard-driven and workstation workflows for the automated rectification of imagery. The IMAGINE AutoSync wizards guide you through the geometric correction process or you can use the workstation for customizable control of your workflow. This chapter explains the steps for edge matching two images using the Edge Matching Wizard and how to use the IMAGINE AutoSync Workstation to georeference a raw image.
All of the data used in this chapter are in the <ERDAS_Data_Home>/examples directory.
Using the Edge Matching Wizard
In this section, you use the IMAGINE AutoSync Edge Matching wizard to align two images so that features in the overlapping area match up. Table 1: Edge Match Image Data Set
Image Name air-photo-1.img air-photo-2.img
Image Date June 1999 March 1997
Image Size (pixels) 1108 W x 1110 H x 1 band 1098 W x 1096 H x 1 band
Pixel Size 18 feet 18 feet
These data files are air photo images of the Oxford, Ohio area. You must have ERDAS IMAGINE running.
1. Click Toolbox tab > Wizard.
AutoSync Workstation
> Edge Match
Using IMAGINE AutoSync Using IMAGINE AutoSync
5 5
Click here to select images
Using the Input tab
In the Input tab, you will add the images to be edge matched. IMAGINE AutoSync will edge match neighboring images, so input image order in the CellArray is important. 1. In the Input tab, click the Open File icon The Input Images dialog opens. .
Click here to open the file Click here to select the file Preview window
2. In the Input Images dialog under File name, select air-photo-1.img from the file list. 3. Click OK in the Input Images dialog. The file air-photo-1.img displays in the Input Images column in the Edge Matching Wizard dialog.
Input image
4. Repeat step 1. through step 3. for the second image, selecting air-photo-2.img this time.
5. Click Next> to continue to the APM Strategy tab in the Edge Matching Wizard.
Using the APM Strategy tab
In the APM Strategy tab, you can adjust the algorithm settings that control the placement of automatically generated tie points in your images. You can also select which input image layer to use to achieve a better point matching result.
Before using APM, confirm that the Initial Pyramid Layer Number is set to 1 in Image Files (General) category of the Preference Editor. This retains the largest pyramid layer when computing pyramid layers to ensure point matching accuracy.
Make sure Defined Pattern is selected
1. Accept the default settings in the APM Strategy tab. Make sure that Defined Pattern is selected. 2. Click Next> to continue to the Edge Match Strategy tab in the Edge Matching Wizard.
Using the Edge Match Strategy tab
In the Edge Match Strategy tab, you can select a refinement method and choose to apply the refinement to the overlapping area only or the whole image.
Click to select Linear Rubber Sheeting
Make sure the buffer size is 180
1. In the Edge Match Strategy tab, click the Refinement Method list and select Linear Rubber Sheeting. 2. Accept the Apply Refinement to default of Overlapping Area Only to apply refinement only to the overlapping area between the images. 3. In the Edge Match Strategy tab, in the Buffer Around the Overlapping Area (pixels): field, keep the default of 180. 4. Click Next> to continue to the Projection tab in the Edge Matching Wizard.
Using the Projection tab
In the Projection tab, you can set a projection for your output images. You can set it to the same projection as the corresponding input image or to another specified projection. NOTE: The Output Projection fields will be greyed out in the Projection tab until you select the Resample geocorrection method in the Output tab.
1. In the Projection tab, accept the default Output Projection of Same as Input Image. 2. Click Next> to continue to the Output tab in the Edge Matching Wizard.
Using the Output tab
In the Output tab, you can specify the properties for your output images, including selecting the geocorrection method and specifying names for the output files and summary report.
Click here to select the Resample method
Click here to open the Resample Settings dialog
Click here to open the Output File Names dialog
Enter a summary report name here
1. In the Output tab, select the Resample geocorrection method. 2. Click Resample Settings... in the Output tab.
10
The Resample Settings dialog opens.
For more information on the geocorrection methods, see Resampling vs. Calibration on page 12.
Make sure Cubic Convolution is selected
3. Accept the default settings in the Resample Settings dialog. Make sure the Cubic Convolution resample method is selected. 4. Click OK in the Resample Settings dialog. 5. In the Output tab, click the Set Output File Names... button. The Output File Names dialog opens.
Enter a default file name suffix here
Click here to select a default output directory
6. In the Output File Names dialog, click the File Selector icon select a default output directory of your choice.
to
7. In the Default Output File Name Suffix field, enter a default file name suffix of your choice, or use the default _output. 8. Click OK in the Output File Names dialog. 9. In the Output tab, make sure the Generate Summary Report checkbox is selected and enter a name of your choice for the HTML summary report. You can also click the File Selector icon of your choice. to select a directory
10. In the Output tab, click Save to save the project. A File Selector opens, and you can save the project to a directory of your choice.
11
11. In the Output tab, click Finish to complete the edge matching process. The AutoSync Job status dialog appears, stating the progress of the edge match operation. 12. Click OK in the AutoSync Job status dialog when the operation is finished. NOTE: Edge matching can take several minutes to run, based upon your hardware capabilities and the size of the image files.
Resampling vs. Calibration Resampling Resampling is the process of calculating the file values for the rectified image and creating the new file. All of the raster data layers in the source file are resampled. The output image has as many layers as the input image. ERDAS IMAGINE provides these widely-known resampling algorithms: Nearest Neighbor Bilinear Interpolation Cubic Convolution Bicubic Spline
Calibration Instead of creating a new, rectified image by resampling the original image based on the mathematical model, calibrating an image only saves the mathematical model into the original image as a piece of auxiliary information. Calibration does not generate new images, so when the calibrated image is used, the math model comes into play as needed. For example, if you want to see the calibrated image in its rectified map space in a Viewer, the image can be resampled on the fly based on the math model, by selecting the Orient image to map system option in the Select Layer To Add dialog. A major drawback to image calibration is that the processes involved with the calibrated image is slowed down significantly if the math model is complicated. One minor advantage to image calibration is that it uses less disk space and leaves the images spectral information undisturbed.
12
Display Output Image

1. Open a 2D View. 2. Open a File Selector. 3. Click the Raster Options tab at the top of the Select Layer To Add: dialog. 4. Make sure the Background Transparent option is checked. 5. Click the File tab at the top of the Select Layer To Add: dialog. 6. In the Select Layer To Add: dialog under Filename, select the output images from the directory in which you saved them. 7. Click OK in the Select Layer To Add dialog. The edge matched output images display in the View.
Use the Viewer Swipe Tool
1. To compare the layers, click Home tab > 2. The Viewer Swipe dialog opens.
Swipe.
13
Drag the slide to swipe over the images
Check if desired to turn on Auto Mode to watch the images being swiped at a rate you choose
3. Check Auto Mode in the Viewer Swipe dialog, and type 500 for the Speed. You can watch as the swipe tool slowly works its way over the images allowing you to evaluate the quality. Experiment with both Vertical and Horizontal direction and different speeds.
View Summary Report
Once you have finished edge matching the images, you can view the HTML summary report to review information about the error, tie points, and so forth. 1. In a Windows Explorer window, browse to the directory where you saved the HTML report (in the Output tab). 2. Click to open the HTML file. The summary report opens in a browser window.
You can experiment with selecting different options in the Edge Matching wizard tabs to produce different results. For example, in the Edge Match Strategy tab, select to apply refinement to the Whole Image (instead of the Overlapping Area Only) or change the Buffer Around the Overlapping Area (pixels) number to see the differences in the resulting output images.
Using the AutoSync Workstation
In this section, you use the georeference workflow in the IMAGINE AutoSync Workstation to georeference a raw Landsat TM image of Atlanta, Georgia, using a SPOT panchromatic image of the same area. The raw Landsat TM image does not have any map information and the SPOT image is rectified to the State Plane map projection.
14
Table 2: Georeference Image Data Set

Image Name tmAltanta.img panAtlanta.img Image Date Feb 1993 Feb 1993 Image Size (pixels) 512 W x 512 H x 6 bands 1401 W x 1301 H x 1 band Pixel Size n/a 32.81 feet
This section explains the steps for using the georeference workflow in the workstation to georeference a raw image (an image without any map information). When georeferencing a rectified image, you do not need to manually collect tie points before running APM. To georeference a raw image in the IMAGINE AutoSync Workstation, follow these basic steps: create a new IMAGINE AutoSync project add an input image add an image to reference against the input image collect manual tie points run APM preview the output image improve output image results (if necessary) review the input and reference image map data information set the output image projection resample or calibrate the output image verify the rectification process view the summary report
Create New IMAGINE AutoSync Project
First, create a new IMAGINE AutoSync project. You must have ERDAS IMAGINE running.
1. Click Toolbox tab >
AutoSync Workstation.
2. The IMAGINE AutoSync Workstation Startup dialog opens, as well as the AutoSync Workstation.
15
Click here to create a new IMAGINE AutoSync project
3. On the IMAGINE AutoSync Workstation Startup dialog, select Create a new project. 4. Click OK. The Create New Project dialog opens.
Click to select the Georeference workflow Enter a project name here
Click to select the Resample geocorrection method
Click to open the Resample Settings dialog
Enter a name for your summary report here
5. On the Create New Project dialog, select the Georeference workflow. 6. In the Project File (*.lap) field, enter a project file name of your choice or click the File Selector icon. 7. On the Create New Project dialog, select the Resample geocorrection method. 8. Click the Resample Settings... button. The Resample Settings dialog opens.
16
Make sure Cubic Convolution is selected Click to accept the default resample settings
9. Accept the default settings in the Resample Settings dialog. Make sure the Cubic Convolution resample method is selected.
IMAGINE AutoSync provides these widely-known resampling algorithms: Nearest Neighbor, Bilinear Interpolation, Cubic Convolution, and Bicubic Spline. In some cases you may want to change the Resample Method, but for this chapter, leave it set to Cubic Convolution. 10. Click OK in the Resample Settings dialog. 11. In the Create New Project dialog, in the Default Output Directory: (*) field, click the File Selector icon to select a default output directory of your choice. 12. In the Default Output File Name Suffix field, enter a default file name suffix of your choice, or keep the default _output. 13. In the Create New Project dialog, make sure the Generate Summary Report checkbox is selected. The name of the project in the Project File field defaults as the summary report name, but you can also click the File Selector icon to select a different name and directory of your choice. 14. If you are using SPOT DIMAP for input and you selected an imagery.tif file for the input image file name, you can run APM without manually measured points.
17
Menu bar Workstation toolbar Project Explorer
Viewer panes
GCP toolbar CellArray
Status bar
Add Input Image
After you have created the IMAGINE AutoSync project, the next step is to add the input image you want to georeference. 1. To add an input image to the project, do one of the following: In the IMAGINE AutoSync toolbar, click the Open Input Images icon . Select File -> Add Images -> Input Images... from the menu bar Right-click on the Input Images folder in the Project Explorer Tree View and select Add Input Image...
The Select Images To Open dialog opens. 2. In the Select Images To Open dialog under Filename, click the file tmAtlanta.img. This file is a Landsat TM image of Atlanta that has not been rectified. 3. Click OK in the Select Images To Open dialog. The input image tmAtlanta.img displays in the IMAGINE AutoSync Workstation.
18
The input file displays in the Project Explorer and in the left Viewer panes
Add Reference Image
After you have added an input image, the next step is to add an image to reference against the input image. 1. To add a reference image to the project, do one of the following: In the IMAGINE AutoSync toolbar, click the Open Reference Images icon .
Select File -> Add Images -> Set Reference Image... from the menu bar Right-click on the Reference Image folder in the Project Explorer Tree View and select Set Reference Image....
The Select Images To Open dialog opens. 2. In the Select Images To Open dialog under Filename, click the file panAtlanta.img. This file is a SPOT panchromatic image of Atlanta. This image has been georeferenced to the State Plane map projection. 3. Click OK in the Select Images To Open dialog. The reference image panAtlanta.img displays in the IMAGINE AutoSync Workstation.
19
The reference file displays in the Project Explorer and the right Viewer panes
Collect Manual Tie Points
Once you have loaded the input and reference images in the IMAGINE AutoSync Workstation, you can manually collect tie points.
This step is necessary here because the input image (tmAtlanta.img) is a raw image (without any map information). When using the IMAGINE AutoSync Workstation to georeference images with map information, you do not need to manually collect tie points before running APM and you can skip this step. 1. In the GCP toolbar, click the Create GCP icon .
20
Click to collect tie points
Click to show tie points in the Viewer panes
Click to change tie point color to yellow
2. In the GCP toolbar, click the Show Selected Points icon
3. To make the input image tie points easier to see in the viewer om the left, right-click in the Color column to the right of Point ID in the first row of the CellArray and select the color Yellow. Repeat this for each tie point in the CellArray. 4. To make the reference image tie points easier to see in the viewer on the right, right-click in the Color column to the right of Y Input in the first row of the CellArray and select the color Yellow. Repeat this for each tie point in the CellArray. 5. In the Main View pane of the input image, click a location to collect a tie point. The point you have created is labeled as 1 in the Main View pane and its X and Y inputs are listed in the CellArray. Also notice that the input image icon in the Project Tree View now has a green border since it now has tie points. 6. In the Main View pane of the reference image, click the same location to collect a tie point.
21
Click to create tie points on both the input image and reference image
These are the X and Y coordinates for tie points in the input image (tmAtlanta.img)
These are the X and Y coordinates for tie points in the reference image (panAtlanta.img)
7. Collect at least three manual tie points in both the input and reference images.
You should choose points that are easily identifiable in both images, such as road intersections and landmarks, so that the images match properly. Also, make sure you scatter your tie points around the images so they are not all concentrated in one place. Try to collect tie points that are close to each of the four corners of the images.
If you are using SPOT DIMAP for input and you selected an imagery.tif file for the input image file name, you can run APM without manually measured points.
22
Your project in the IMAGINE AutoSync Workstation should now look similar to the following:
Run APM
After collecting several tie points in the input and reference image, the next step is to run automatic point matching (APM) to automatically generate more control points for your images.
Before using APM, confirm that the Initial Pyramid Layer Number is set to 1 in Image Files (General) category of the Preference Editor. This retains the largest pyramid layer when computing pyramid layers to ensure point matching accuracy. 1. In IMAGINE AutoSync toolbar, do one of the following to run APM: In the IMAGINE AutoSync toolbar, click the Run APM icon Select Process -> Run APM from the menu bar Right-click on the input image (tmAtlanta.img) in the Project Explorer Tree View and select Run APM
23
The Status bar at the bottom of the workstation displays the RMSE and Error Standard Deviation results. The APM points generated populate the CellArray and display in the Viewers for both the input and reference images.
The APM points display in the Viewer panes and populate the CellArray
The Status bar displays the RMSE and error standard deviation results
Preview Output Image
After you run APM, you can preview the output image to make sure you are satisfied with the results before resampling or calibrating. 1. To preview the output image, right-click on the input image (tmAtlanta.img) in the Project Explorer Tree View and select Preview Output. 2. Click Reset Image to Viewer Size icon 3. Click Zoom In by Two icon . .
A preview of the output image displays in the Viewer.
24
Improve Output Image Results
If you preview the output results and the image is warped, shows black images, or produces other unacceptable output, it is most likely the result of incorrect APM tie points or an inappropriate sensor model. In this chapter, if you are dissatisfied with the results, there are most likely incorrect tie points that you should delete before resampling. If the Error Std. Dev. is higher than 2.0, you should also follow these steps to improve the tie points. If you delete incorrect APM points and the results are still poor, then you should try changing the sensor model in the IMAGINE AutoSync Project Properties dialog.
To learn more about improving APM results, see APM Engine on page 4. 1. Right-click on the input image (tmAtlanta.img) in the Project Explorer Tree View and select Review Points. The input and reference images display in the Viewer panes, showing the tie points.
25
Enter 2 for the error threshold search criteria
Click to select tie points in the CellArray with an error threshold of 2
2. In the GCP toolbar, enter or use the nudgers to the right of the field to enter 2 in the error threshold text box. 3. In the GCP toolbar, click Select GCPs with Error Threshold icon .
The tie points with an error higher than 2 are highlighted in the CellArray. Note the Total Points value and the Selected value, located to the right of the Select GCPs with Error Threshold icon. Total Points - total number of manual GCPs and APM GCPs. Selected - number of GCPs you have selected
26
Click here to drive through the selected points in the CellArray
4. Click the Drive To icon CellArray.
to click through the selected points in the
As you click through the points, the points in the Viewers will be highlighted with a box. 5. When you find a point with a high error, click Delete GCP icon The selected point is deleted from the Viewers and the CellArray. 6. After you delete all the points with a high error, right-click on the input image (tmAtlanta.img) in the Project Explorer Tree View and select Preview Output to see the current output. 7. Right-click on the input image (tmAtlanta.img) in the Project Explorer Tree View and select Review Points to return to the other view. .
Review Image Map Data
The next step is to review the image map data for the input and reference image. You can review the map data to learn about the map and projection information in order to determine if you want the output image to have the same projection as the reference image.
27
1. Click on the reference image in the Main View pane to select it.
Click to open the ImageInfo dialog
Click to select panAtlanta.img
2. On the IMAGINE AutoSync toolbar, click the ImageInfo icon The ImageInfo dialog for panAtlanta.img opens.
Note map and projection information
Note the information in the Map Info section and that the Projection Info section shows that the map is georeferenced to State Plane. 3. When you are finished, select File -> Close in the ImageInfo dialog.
28
4. Click on the input image in the Main View pane to select it.
Click to open the ImageInfo dialog
Click to select tmAtlanta.img
5. On the IMAGINE AutoSync toolbar, click the ImageInfo icon 6. The ImageInfo dialog for tmAtlanta.img opens.
Note map and projection information
Note the information in the Map Info section and that the Projection Info section shows that this is a raw image with no projection information. Therefore, for this chapter, use the input projection from the reference image (panAtlanta.img) for the output image.
29
Set Output Image Projection
Since the image to be georeferenced is a raw image with no projection information, you need to change the project properties to set the output projection before resampling or calibrating the image. 1. To change the output projection, do one of the following: Select Process -> Project Properties... from the menu bar In the IMAGINE AutoSync toolbar, click the Edit Project Properties icon The IMAGINE AutoSync Project Properties dialog opens. 2. Click the Projection tab. The Projection tab opens in the IMAGINE AutoSync Project Properties dialog. 3. In the Output Projection section, select Same as Reference Image. The projection information from the reference image displays (greyed out).
Click here to use the same projection as the reference image The reference image projection info displays here
4. Click the OK button to close the IMAGINE AutoSync Project Properties dialog.
Resample Output Image
Resampling is the process of calculating the file values for the rectified image and creating the output file. All of the raster data layers in the source file are resampled. The output image will have as many layers as the input image.
30
1. To resample the output image, do one of the following: Select Process -> Calibrate/Resample from the menu bar Right-click on the input image (tmAtlanta.img) in the Project Explorer Tree View and select Calibrate/Resample
You can change the resample settings in the Output tab in the IMAGINE AutoSync Project Properties dialog. You may want to experiment with changing the resample settings later, but this chapter uses the default settings. The resampled output image (tmAtlanta_output.img) displays in the workstation viewer and the output image name now displays in the Output Images folder in the Project Explorer. Your project in the IMAGINE AutoSync Workstation should now look similar to the following:
31
The resampled output image displays in the Viewer and Project Tree View
Verify Output Image
Once the output image is created, you can use the Workstation to perform the output image verification. You can verify that the input image (tmAtlanta.img) has been correctly georeferenced to the reference image (panAtlanta.img) by visually checking that they conform to each other using the Viewer Blend/Fade, Viewer Swipe, or Viewer Flicker verification tools.
Use the Viewer Blend/Fade Tool
1. To perform visual verification using the Viewer Blend/Fade tool, click the Start Blend Tool icon on the IMAGINE AutoSync toolbar.
The Viewer Blend/Fade dialog opens.
32
Drag the slide to blend and fade the images Check if desired to turn on Auto Mode to watch the images being blended together
2. Select Auto Mode in the Viewer Blend/Fade dialog, and type 500 for the Speed. You can watch as the tool slowly blends the images, allowing you to evaluate the quality. Experiment with both different speeds or use the slide to blend and fade the images.
Use the Viewer Swipe Tool
1. To perform visual verification using the Viewer Swipe tool, click the Start Swipe Tool icon on the IMAGINE AutoSync toolbar.
The Viewer Swipe dialog opens.
Drag the slide to swipe over the images
Check if desired to turn on Auto Mode to watch the images being swiped at a rate you choose
33
2. Check Auto Mode in the Viewer Swipe dialog, and type 500 for the Speed. You can watch as the swipe tool slowly works its way over the images allowing you to evaluate the quality. Experiment with both Vertical and Horizontal direction and different speeds.
Use the Viewer Flicker Tool
1. To perform visual verification using the Viewer Flicker tool, click the Start Flicker Tool icon on the IMAGINE AutoSync toolbar.
The Viewer Flicker dialog opens.
Click to switch between the images
Check if desired to turn on Auto Mode to watch the images switch from top to bottom at a rate you choose
2. Check Auto Mode in the Viewer Flicker dialog, and type 500 for the Speed. You can watch as the flicker tool switches between the top and bottom images, allowing you to evaluate the quality. You can also click Manual Flicker to quickly switch between the images. Experiment with different speeds.
View Summary Report
You can view the summary HTML report for more information.review information about the error, tie points, and so forth. 1. To view the summary report, do one of the following: In the GCP toolbar, click the Summary Report icon Right-click on input image (tmAtlanta.img) in the Project Explorer Tree View and select Review Report. Open a Windows Explorer window, browse to the directory where you saved the HTML report (in the Create New Project dialog) and open the .html file
The summary report opens in a separate browser window.
34
35
36
37
38
Index
APM, Automatic Point Matching 4, 41 AutoSync Workstation 32 Create New Project dialog 34 GCP toolbar 38 Georeference workflow steps 33 Project Properties dialog 48 Resample Settings dialog 34 Viewer Blend/Fade tool 50 Viewer Flicker tool 51 Viewer Swipe tool 50 Workstation Startup dialog 33
I
Image information 45
M
Map information 45 Map projection State Plane 32 Map projection, set 48
N P
Nearest Neighbor 30, 35
Polynomial transformation 14
R
Rational Polynomial Coefficients 15 Raw imagery workflow 20 Resample 30, 48 Bicubic Spline 30, 35 Bilinear Interpolation 30, 35 Cubic Convolution 30, 35 Nearest Neighbor 30, 35 Review points 43 Rigorous Orbital Pushbroom transformation 15 RMSE analysis 18 RMSE results 41 Rubber sheeting transformation 14
Bicubic Spline 30, 35 Bilinear Interpolation 30, 35
C
Calibration 30 Cubic Convolution 30, 35
D
Data air photos 23 Landsat multispectral 32 raw 32 SPOT panchromatic 32 Direct Linear Transform 16
S T
E
Edge Matching Wizard 23 APM Strategy tab 25 Edge Match Strategy tab 26 Input tab 24 Output File Names dialog 29 Output tab 28 Projection tab 27 Refinement method list 27 Resample settings dialog 29 Edge Matching workflow 20 Error standard deviation results 41, 43 Example data vii
Summary report 32, 35, 52
Tie points, analysis 18 Tie points, collect 38 Tie points, generate automatically 4, 41
V
Verify Output image 49 Viewer Blend/Fade tool 50 Viewer Flicker tool 51 Viewer Swipe tool 31, 50
F
File extension (*.lap) 34
Georeference workflow 20, 32
Index
57
58
Index

AutoSync PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

AutoSync PDF

Transféré par

Droits d'auteur :

Formats disponibles

IMAGINE AutoSync

Users Guide December 2010

Copyright 2010 ERDAS, Inc.

Introduction to IMAGINE AutoSync . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Table of Contents Table of Contents

Using IMAGINE AutoSync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

List of Tables List of Tables

Conventions Used in This Book

These paragraphs provide software-specific information.

These paragraphs contain important tips.

There are several ways to open Help:

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

Selecting Input and Reference Images

Introduction to IMAGINE AutoSync

Digital Elevation Model (DEM)

APM Strategy Parameters

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

Advanced Point Matching Strategy Dialog

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

Ideal Situations for Good APM Performance

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

APM Trouble Shooting and Tips

Correlation Size and Least Squares Size Intended Number of Points

Introduction to IMAGINE AutoSync

Intended Number of Points Column Increment and Line

Intended Number of Points Column Increment and Line

Introduction to IMAGINE AutoSync

Image to Image (2D) Transforms

Introduction to IMAGINE AutoSync

Ground to Image (3D) Transforms

Approximate Sensor Models

Introduction to IMAGINE AutoSync

Rational Polynomial Coefficients (RPC)

Modeling TroubleShooting and Tips

Introduction to IMAGINE AutoSync

Number of Tie Points

Introduction to IMAGINE AutoSync

IMAGINE AutoSync Tips and Hints

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

Using the IMAGINE AutoSync Workstation

Using the IMAGINE AutoSync Wizards

Introduction to IMAGINE AutoSync

IMAGINE AutoSync Workflows

General IMAGINE AutoSync Tips and Hints

Introduction to IMAGINE AutoSync

Introduction to IMAGINE AutoSync

Refer to Data Preparation on page 2.

Refer to APM Engine on page 4.

Refer to Modeling on page 13.

Refer to IMAGINE AutoSync Tips and Hints on page 17.

Refer to Modeling on page 13.

Introduction to IMAGINE AutoSync

Using IMAGINE AutoSync

Using the Edge Matching Wizard

Image Name air-photo-1.img air-photo-2.img