Vous êtes sur la page 1sur 4

International Journal of Computer Trends and Technology- volume3Issue1- 2012

Motion Object Detector with Higher Detection Levels


BapiRaju Mudunuri1, D.V.Ravi Kumar2
1

M.Tech CSE, Sri Vasavi Engineering College, Pedatadepalli, Tadepalligudem. W.G.Dt., A.P. India.

Asst. Professor, Dept of CSE, Sri Vasavi Engineering College, Pedatadepalli, Tadepalligudem. W.G.Dt. A.P., India.

Abstract This system is all about motion object detection. The system detects the moving objects w ith a very good accuracy and precision levels. A w eb camera is used to capture the live video and the motion detection algorithm takes this as input and does its w ork and finally the output is displayed as live video show ing or indicating the moving objects w ith a red solid marker. It can be used as a video surveillance system. The syste m is tested at various lightning and environmental conditions. The output of the tests has show n that it is able to detect the moving objects very efficiently. The accuracy and precision levels of the syste m w as found to be much better than the current ex isting motion detection and tracking systems. Keywords video surveillance, Motion detection, object trac king.

moving object at that place and thus everything is static over there. During pixel comparison, an appropriate threshold value is to be chosen so that the motion detector system gives out apt results. This is all about the base technique used for detecting the moving objects. But currently, many motion detectors are employing certain artificial intelligence techniques are used along with the base technique in order to build a much more accurate motion detector system. These systems have the capability of detecting particular objects of interest by using preloaded database of such objects. This is all about the introductory concept related to motion detection and tracking systems. 2. RELEVANT WORK We observe and analyze various currently used techniques for building the motion detection and tracking systems. Analysis is made on each of the techniques along with the advantages and disadvantages. Hanzi Wang et al [1] proposed a motion detection method using an effective appearance filter. It uses a similarity measure based on spatial color mixture of Gaussians (SMOG) for particle filters. The color histogram and the spatial color layout are considered in this method. Since color information is extensively used in this method, it builds a better motion detector. The base technique of frame comparison is clearly described in the motion and feature based person tracking system [3] by Augustin et al. It also describes ho w the background subtraction technique would be useful in eliminating the real time problems occurring in the motion detection systems. Stauffer et al [2] proposed the method of pixel modeling using Gaussians for motion detection. Each pixel is classified based on the Gaussian distribution along with the consideration of the background model. The contour extraction of moving objects [4] proposed by Lili Qiu et al, says that fusing information from the color segmentation, motion segmentation and active contour

1. INTRODUCTION Detecting moving objects and accurately tracking them is useful in a wide range of applications like video surveillance, security systems, etc. Depending on the application, the system should have the accuracy and precision levels. For example if it is a security system then the accuracy and precision levels need to be very high. And for simpler systems minimal accuracy is essential. There are many techniques available for motion detection and tracking systems. For better ones several technologies can be clubbed with each other which leads to a better accuracy in the overall system. Software systems can be clubbed with electronic radar systems to build efficient motion detection and object tracking systems. The commonly used technique in most of the motion detection and tracking systems is to go for the corresponding pixel comparison of two successive frames of the video input given through a webcam. If a change in the corresponding pixels is found, then it is confirmed that an object is being moved at that place. And if there is no change in the pixels, then it is confirmed that there is no

ISSN: 2231-2803 http://www.internationaljournalssrg.org

Page 192

International Journal of Computer Trends and Technology- volume3Issue1- 2012


leads to an accurate extraction of the boundary of the moving objects. It works well not only for a single moving object but also for many. The moving target classification [5] proposed by Lipton et al describes the end to end method for extracting moving targets and classifying them in to predefined categories according to image based properties. Wren et al [6] proposed the Pfinder method for tracking and interpretation of people. This system uses a multi-class statistical model of color and shape to obtain a 2-D representation of head and hands in a wide range of viewing conditions. The flow and organization of this review paper has followed four very thorough, excellent surveys conducted by (Ko, 2008; Wang et al., 2003; Hu et al., 2004; Kumar et al., 2008) when discussing the general framework of automated visual surveillance systems as shown in Fig. 1, enriching with the general architecture of a video understanding system (Bremond et al., 2006) in behavior analysis and with expandable network system architecture as illustrated in (Cohen et al., 2008). The main intent of this paper is to give engineers, scientists, and/or managers alike, a high-level, general understanding of both the theoretical and practical perspectives involved with a visual surveillance system and its potential challenges while considering implementing or integrating a visual surveillance system[9]. lightning scene lightning can cause problems for many background models. 3. PROPOSED METHODOLOGY Our proposed system aims at tracking the moving objects more accurately. We detect and track the exact boundary line of the moving object. We use the base technique of comparing the corresponding pixels of two successive frames. And whenever a change in the pixels is found then we confirm that motion of some object is being taking place. A web cam is used to capture the live video and our proposed motion detection algorithm is applied on this video at real time. Finally the output is displayed as the video containing red color solid marking indicating that the object is moving. We use filters from the Aforge.net framework which is an open source one. The structural design of our proposed method shown in Fig.1

Fig. 1 System architecture As shown in the figure 1, our proposed system consists of three steps. A detailed explanation of each step is given belo w. 1.INPUT WEBCAM VIDEO In this step, live video is captured using a webcam and fed further for motion object detection processing. The camera involved may be a simpler one or a very advanced one having a very high resolution. This is an important factor to be considered because the overall accuracy and precision of our proposed system depends on the resolution of the camera being used. And also the light sensitivity of the camera is also an important factor to be considered. To accomplish this step we use the d sho w from Microsoft. It has got a very rich set of api that can be used to build video related applications. 2. PROCESS MOTION DETECTION

Fig. 1. A general framework of an automated visual surveillance system

Friedman and Russell [7] proposed a method to explicitly classify the pixel values into a few distributions corresponding to various factors. With this technique, they are successful in handling the shadows in the real time captured video. Ridder [8] has recently implemented the system using Kalman Filter that has made it more robust to

ISSN: 2231-2803 http://www.internationaljournalssrg.org

Page 193

International Journal of Computer Trends and Technology- volume3Issue1- 2012


In this step, the actual core technique of motion detection comes in to action. Initially, the difference between the successive frames is found out. This is accomplished with a Difference filter. The threshold level of the difference is to be a very balanced value. For this, we use another filter called the Threshold filter. Due to various undesirable environmental conditions like lightning and false movements of the camera many false positives of the output are obtained. In order to avoid this we apply another filter called the Erosion filter. After the application of the erosion filter it is ensured that false positives of the results are eliminated. The application of filters is as shown in the Fig. 2.

Fig. 3 Hand movement detected Fig. 3 shows how the system is able to detect the movement of the hand. Observe the red color solid marker indicating the movement of the hand. Whatever be the speed of the hand movement the system successfully detects the motion as shown in the above figure. We have observed that if the hand is moved very slo wly, then the threshold value of the filter must be aptly set. TEST: 2

Fig. 2 Applying Filters. 3.DISPLAY MOTION OUTPUT This is the final stage of the system. It displays the output after the motion detection process is applied on the live video. The output display shows the live video along with the solid red color marking indicating that the motion is being taking place. The display output accurately detects and tracks the exact boundary line of the moving object. The red color solid marker is chosen as it would clearly highlight the object moving. 4. EXPERIMENTAL RESULTS The system is tested and the output is presented as sho wn below. Observe the red color solid marker indicating the motion of some object. The test is conducted under standard indoor lightning conditions. The accuracy of the system highly depends on the environmental lightning conditions. TEST: 1 Fig. 4 Person moving sideways In Fig. 4, the person moves sideways. One can clearl y observe that the system is able to accurately detect the actual boundary line of movement of the person. Here, when the motion is detected and the same is indicated and also the internal detail of the moving object is still retained. The system is not constrained to detect only one moving object in the scene but also multiple moving objects can be detected. The threshold filter used in the system is an important factor. T he value of the threshold filter decides the accuracy in detecting and tracking the exact boundary line of the moving object. We have observed that by varying the threshold value of the threshold filter the red color solid marker that is surrounding the head of the person in Fig. 4 has also varied. By lowering the threshold value, the head of the person is much more accurately detected and tracked TEST: 3

ISSN: 2231-2803 http://www.internationaljournalssrg.org

Page 194

International Journal of Computer Trends and Technology- volume3Issue1- 2012


system works better than other motion detection and tracking systems of its class. REFERENCES [1] Wang, H., et al., " Adaptive object tracking based on an effective appearance filter". IEEE T ransactions on Pattern Analysis and Machine Intelligence,(2007). [2] C.Stauffer and W.E. L.Grimson, Adaptive background mixture models for real-time tracking, in Proc. IEEE Conf. Computer Vision and Pattern Recognition, 1999. [3] Augustin , Motion and Feature Based Person Tracking In Surveillance Videos in Proc. IEEE Conf. International Conference on Emerging Trends in Electrical and Computer Technology, 2011. [4] L. Qiu and L.Li,Contour extraction of moving objects, in Proc. IEEE Intl Conf. Pattern Recognition, vol. 2, 1998, pp. 14271432. [5] A. Lipton, H. Fujiyoshi and R.Patil,Moving target classification and tracking from real-time video, In DARPA Image Understanding Workshop, November 1998. [6] C.Wren, A. Azarbayejani, T . Darrell, A. Pentl, Pfinder: Real-time tracking of the human body,IEEE TransPattern Analysis and Machine Intelligence. [7] NirFried man and Stuart Russel Image Segmentation in Video sequences: A probabilistic approach in Proc of the Thirteenth Conference on Uncertainity in Artificial Intelligence(UAI),1997. [8] Christof Ridder,Olaf Munkelt and Harald Kirchner. Adaptive Background Estimation and Foreground detection using Kalman-Filtering Proc of Intenational Conf on Recent Advances in Mechatronics, 1995. [9] Ko, T. (2008). A Survey on behavior analysis in video surveillance for homeland security application, AIPR, pp. 1 8, 37th IEEE Applied Imagery Pattern Recognition Workshop.

Fig. 5 Person walking Fig. 5 shows how the motion of a person walking is detected. And in this case too, the internal detail of the person who is moving is retained. We have observed that when these kind of large objects are to be tracked, then the resolution of the camera used must be pretty high. If this system is used in outdoor, the accuracy may be further enhanced because in this case the lightning conditions would be high. The above tests are made under standard indoor lightning conditions. We have found from the above results that the size of the object that is moving has a very close relationship with the resolution of the camera that is used in the system. For detecting and tracking moving objects of very small size, the camera used must be of a very high resolution one. 5. FUTURE WORK This system can be further enhanced by clubbing it with other electronic based technologies. For example, it can be clubbed with radar systems so that the two technologies can work in parallel and the motion of the moving object can be detected and tracked with much more accuracy and precision levels. Additional particle filters can be used too to make it better. Certain artificial intelligence techniques can be clubbed with this system so that it can be used in artificial intelligence related applications like robotics. Several other technologies can be mixed too. 6. CONCLUSION The system performs better in a wide range of environmental conditions. Whatever be the lightning conditions the system successfully detects and tracks the moving objects in real time video applications. And also the system is robust enough to handle cameras of various frame rates. However, experimental results have shown that the accuracy and precision levels of the system depends on the frame rate of the camera being used. If the processing power of the underlying system is high, then our

ISSN: 2231-2803 http://www.internationaljournalssrg.org

Page 195

Vous aimerez peut-être aussi