Académique Documents
Professionnel Documents
Culture Documents
Abstract— Pavement crack detection from images is a challeng- significant transportation system failure (e.g. airport run-
ing problem due to intensity inhomogeneity, topology complexity, ways or bridge decks), and need to be detected accurately
low contrast, and noisy texture background. Traditional learning- and timely. Therefore, it is necessary to develop an automatic
based approaches have difficulties in obtaining representative
training samples. We propose a new unsupervised multi-scale crack detection method for pavement inspection. However,
fusion crack detection (MFCD) algorithm that does not require automatic crack detection is challenging due to intensity inho-
training data. First, we develop a windowed minimal intensity mogeneity, topology complexity, low contrast, and noisy back-
path-based method to extract the candidate cracks in the image grounds [4], [5]. Since a pavement crack is usually thin and
at each scale. Second, we find the crack correspondences across long, the most distinguishable characteristic is its continuous
different scales. Finally, we develop a crack evaluation model
based on a multivariate statistical hypothesis test. Our approach property. As a result, local image-processing methods, such as
successfully combines strengths from both the large-scale detec- intensity thresholding [5], [6], and edge detection based meth-
tion (robust but poor in localization) and the small-scale detec- ods [7], [8] can only obtain a set of disjoint crack fragments
tion (detail-preserving but sensitive to clutter). We analyze and with a high false positive rate. Machine learning approaches
experimentally test the computational complexity of our MFCD have been proposed, but the selection of parameters depends
algorithm. We have implemented the algorithm and have it
extensively tested on three public data sets, including two public on pavement appearance, crack variations, and image quality.
pavement data sets and an airport runway data set. Compared Moreover, the results from these learning methods depend on
with six existing methods, experimental results show that our the quality of manually-labeled training data sets. It is labor-
method outperforms all counterparts. Specifically, it increases expensive and not easy to obtain the representative training
the precision, recall, and F1-measure over the state-of-the-art by data, especially in the real applications, because crack images
22%, 12%, and 19%, respectively, on one public data set.
have large variations due to different lighting conditions, sur-
Index Terms— Crack detection, multi-scale image fusion, face types, and background texture. Morphological methods,
pavement inspection, robotic airport runway inspection. such as the seed-based approaches [9], [10] and minimal path
selection methods [11]–[13], have been successfully used in
I. I NTRODUCTION pavement crack detection by exploiting the connectivity among
crack pixels. However, their performance usually depends
C RACKS are common pavement surface distresses that
affect road performance and periodical road surveys
are necessary to assess pavement surface conditions [1]–[3].
on the parameter choice which requires manual extensive
parameter tuning for each dataset.
In fact, crack images exhibit different characteristics at
Traditional manual crack detection methods are very time-
different scales: at a large scale, crack detection is reliable, but
consuming, labor-intensive, with low-accuracy, and error
its localization is poor and may miss thin cracks; at a small
prone. Even thin cracks may be early warning signs of
scale, details are preserved, but detection suffers greatly from
Manuscript received November 8, 2017; revised May 14, 2018; accepted clutters in background texture. However, the key challenge is
July 12, 2018. This work was supported in part by the National Sci- how to effectively combine the strengths of different scales
ence Foundation under Grant NRI-1426752, Grant NRI-1526200, and Grant to improve crack detection performance. To deal with the
NRI-1748161, in part by the National Science Foundation of China under
Grant 61305107, in part by the Industrial Robot Application of Fujian challenges, we propose a new unsupervised multi-scale fusion
University Engineering Research Center, Minjiang University, under Grant based crack detection (MFCD) algorithm that does not require
MJUKF-IRA201803 and Grant MJUKF-IRA201807, in part by the Open Fund manually-labeled training data (See Fig. 1). MFCD computes
Project of Fujian Provincial Key Laboratory of Information Processing and
Intelligent Control, Minjiang University, under Grant MJUKF201732, in part the maximum average score of cracks at different scales.
by the Fundamental Research Funds for the Central Universities under Grant By extracting crack features for every probable target scale
3122016B006, and in part by the Chinese Scholarship Council. The Associate and evaluating the cracks jointly across scales, MFCD fuses
Editor for this paper was H. G. Jung. (Corresponding author: Haifeng Li.)
H. Li and Y. Liu are with the CS Department, Civil Aviation University of and filters cracks at all scales.
China, Tianjin 300300, China, and also with the Fujian Provincial Key Labo- We have implemented our MFCD algorithm and extensively
ratory of Information Processing and Intelligent Control, Minjiang University, tested it on three datasets in comparison to six existing meth-
Fuzhou 350108, China (e-mail: hfli@cauc.edu.cn).
D. Song and B. Li are with the CSE Department, Texas A&M Uni- ods. Experimental results show that our method consistently
versity, College Station, TX 77843 USA (e-mail: dzsong@cse.tamu.edu; outperforms the counterparts. More specifically, our MFCD
binbinli@tamu.edu). algorithm increases Precision, Recall and F1-measure over the
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. state-of-the-art by 22%, 12% and 19%, respectively, on one
Digital Object Identifier 10.1109/TITS.2018.2856928 public dataset.
1524-9050 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Fig. 2. An illustration of our windowed minimal intensity path based crack detection method. (a) Original pavement image. (b) Crack seed extraction results.
(c) Crack seed clustering and filtering illustration. (d) WMIP generation result. (e) Path verification results. (f) Crack region detection by path growing.
path generation, and designs the windowed path generation algorithm as illustrated in Fig. 1. First, we generate the multi-
algorithm to decease the computing time. scale images by Gaussian blurring with different kernel sizes.
Our MFCD algorithm is also inspired by multi-scale analy- Second, we find the candidate cracks at each scale by utilizing
sis methods which capture the intrinsic geometrical structure a windowed minimal intensity path selection based method.
that is key in human visual perception. Existing techniques, Finally, we determine the cracks correspondence across dif-
such as wavelet transforms [28], particle filters [29], beamlet ferent scales and propose a statistical crack evaluation model
transform [30], are essential different types of multi-scale to compute the average score of each crack at different scales
methods. The main challenge of these multi-scale analysis for crack fusion.
methods is to select the right scale for identifying useful
features or how to combine the detections at different scales
A. Multi-Scale Images Generation
to form output. Most approaches take a simplistic cue com-
bination: they either accept results (after thresholding) at all Using Gaussian blurring technique in computer vision [31],
scales, or accept results that appear at the coarsest scale. Our the image at the s-th scale level, I s , can be produced from the
method fuses results from different scales using a statistical convolution of a variable-scale Gaussian, G(u, v, σs ), with an
hypothesis test. input image I 0 .
1 −(u 2 +v 2 )/2σs2
III. P ROBLEM F ORMULATION G(u, v, σs ) = e . (1)
2πσs2
A. Assumption
Thus, a set of images with different scales are obtained with
We assume a crack has a lower intensity value than that different σs , which serve as the input to the following steps.
of the background image. This assumption can be accepted
in most general cases since cracks always absorb more light
than other areas and often appear as dark curves or tapes in B. Candidate Crack Extraction at Each Scale
the image. We also assume that pixels belonging to the same As shown in Fig. 2, we propose a Windowed Minimal
crack form a continuous path with an arbitrary shape. Intensity Path (WMIP) method to find candidate cracks at each
scale.
B. Notations 1) Crack Seed Extraction (Fig. 2(b)): Let g(x) be the
intensity value of pixel x, and Te be a threshold. We divide
Common notations are defined as follows,
the whole image I s into grid cells, denoted as Yis , i =
• I s , s = 0, 1, . . . , n s − 1, the digital image at the s-th
1, 2, . . . , n y . Each cell is a set of m × m pixels. In each grid
scale with I 0 denoting the original image. All images are
cell Yis , we select a pixel xis as a crack seed when the following
in grayscale.
two conditions are satisfied: 1) xis is the darkest pixel in Yis ,
• xis = [u, v]T , the i -th pixel in I s , with (u, v) being image
and 2) xis is within the top Te percent darkest pixels in I s .
coordinates.
• C s , the pixel set for the candidate cracks in I s . xis = argmin g(x),
• C ∗ , the detected crack pixel set as the algorithm output. x
s.t. x ∈ Yis , g(x) < ge , (2)
C. Problem Definition where ge is the pixel intensity value of the top Te percent
Our ultimate goal is to extract all cracks from an input pave- darkest pixels in I s .
ment image by multi-scale crack fusion, as shown in Fig. 1. As the output of the step, let us denote E s as the set of
Thus, our crack detection problem is defined as follows, crack seeds.
Definition 1 (Crack Detection): Given I 0 , generate multi- 2) Crack Seed Clustering and Filtering (see Fig. 2(c)):
scale images I s , s = 0, 1, . . . , n s − 1, extract the candidate With E s obtained, we use DBSCAN algorithm [32] to group
cracks C s from each I s , then fuse C s to obtain C ∗ . crack seeds into clusters, meanwhile, find the isolated crack
seeds that need to be removed. DBSCAN is a density-based
IV. M ULTI -S CALE PAVEMENT C RACK D ETECTION clustering algorithm, which can group together points that are
Pavement cracks in an image tend to have one or more closely packed and identify isolated outliers in low-density
particular salient scales. To combine the strengths of multiple regions. The advantages of DBSCAN include: 1) it does not
scales, we propose a three-step multi-scale crack detection require users to specify the number of clusters in the data
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Since f (Ci ) is the sum of squares of multiple normal Algorithm 1 WMIP Based Crack Detection
distributions according to (15), f (Ci ) follows χ 2 distribution input : I s
with n s degrees of freedom. Define F(x, n s ) be the cumulative output: C s
distribution function of χ 2 distribution with n s degrees of
1 Detect all crack seeds from I s to generate E s ; O(κ)
freedom, the probability of a value from χ 2 distribution larger
than x is 2 Cluster and filter E s using DBSCAN; O(κ log κ)
T 2 log T p
P{ f (Ci ) ≥ x} = 1 − F(x, n s ). (16) 3 foreach xis ∈ E s and xsj ∈ E s do O( p m 4 κ 2 )
4 if x j ∈ Di then
s s
By setting the significance level as α, we can obtain
5 Compute Pi,s j by solving (5) using Dijkstra
P{ f (Ci ) ≥ x} = α. (17) algorithm; O(T p2 log T p )
Since the function F is continuous and strictly monotoni- 6 Compute mean intensity m(Pi,s j ) using (6); O(T p2 )
cally increasing, combining (16) and (17), we can obtain
7 if m(Pi,s j ) < gm then
x = F −1 (1 − α, n s ), (18) 8 Discard Pi,s j ; O(1)
where F −1 (·) is the inverse function of F(·). Thus, we reject
9 foreach Pi,s j ∈ P s do O(κ log κ)
H0 if
10 repeat O(κ log κ)
f (Ci ) ≤ F −1 (1 − α, n s ). (19) 11 Compute Gis using (7); O(n i )
12 Cis = Cis ∪ Gis ; O(1)
After the statistical hypothesis test, we remove the false-
13 until Cis is unchanged;
positive cracks. Combine all the cracks remained to obtain C ∗ .
14 foreach Cis and C sj do O(|Cis ||C sj |)
V. A LGORITHM A NALYSIS 15 Compute di, j using (9); O(|Cis ||C sj |)
A. Complexity Analysis 16 if di, j < Tg then
We summarize the proposed WMIP based candidate crack 17 Merge Cis and C sj into the same group;
detection method in Algorithm 1 to facilitate our analysis. O(|Cis | + |C sj |)
Recall κ is the total amount of pixels in I s . For crack seed
18 foreach Cis ∈ C s do O(|Cis |)
extraction, since at most κ/m 2 cells are obtained, with each
19 if |Cis | < Ts then
cell containing m 2 pixels, the crack seeds can be detected in
20 Remove Cis from C s ; O(|Cis |)
O(κ/m 2 × m 2 ) = O(κ) time. Crack seed clustering takes
O(|E s | log |E s |) [32]. Obviously, |E s | ≤ κ/m 2 ≤ κ, thus, 21 return C s ; O(1)
crack seed cluster has a time complexity O(κ log κ). Since
we only find the WMIP within a rectangular window Wi,s j
whose diagonal length is smaller than T p , the maximum size
be considered as a constant. Thus, crack matching across two
of Wi,s j is 12 T p2 , thus, when proceeding the Dijkstra algorithm, s
scales has a time complexity O(|Cisi ||C j j |). Computing all
the number of vertex is smaller than 12 T p2 , and the number crack correspondences across each pair of adjacent scales takes
of edges is no more than 4T p2 . So the time complexity of s s
O(n s |Ci i ||C j j |) time. When computing the average score of
each WMIP generation is O(T p2 log T p ). There are at most any crack Cis , the scores of |Cis | pixels in Cis need to be cal-
κ 2 /m 4 paths in all because |E s | ≤ κ/m 2 . Thus, the time culated firstly, as shown in Eq. (13). The neighboring window
T 2 log T p
complexity for all WMIPs generation is p m 4 κ 2 . Since the for each pixel is (3σ p + 1) × (3σ p + 1). Thus, the overall
length of Pi,s j is no more than the size of Wi,s j , we have computation of average score of Cis takes O((3σ p + 1)2 |Cis |)
|Pi,s j | ≤ 12 T p2 . Therefore, we can compute m(Pi,s j ) in O(T p2 ) time. Since σ p is a constant, computing ρ(Cis ) has a time
time. Crack region detection by path growing takes O(κ log κ) complexity O(|Cis |). Since the total number of crack pixels
time. Computing di, j between two sets with size |Cis | and |C sj | in each I s is smaller than κ, merging cracks across n s scales
needs O(|Cis ||C sj |) time. Usually, the total number of crack takes in O(n s κ) time. Define the total number of Ci be n.
regions is very small and can be considered as constant. Thus, Considering |Ci | ≤ κ, the time complexity of hypothesis test
the time complexity of crack grouping is also O(|Cis ||C sj |). for all merged cracks is O(nκ). Thus, the crack selection and
s s
verification for each two scales takes in O(max(n i i , n j j )) time.
To summarize, since |Cis ||C sj | < κ 2 , the most computationally
expensive step in Algorithm 1 is WMIP generation using Since both |Cis | and n are much smaller than κ, the computa-
T p2 log T p 2
Dijkstra algorithm. Thus, the computational complexity of our tional complexity of our MFCD algorithm is O(n s m4
κ ).
T 2 log T p
WMIP based crack detection algorithm is O( p m 4 κ 2 ). Theorem 2: Our MFCD algorithm runs in
T p2 log T p 2
O(n s m 4 κ )
Theorem 1: The computational complexity of the proposed time.
T p2 log T p
WMIP based crack detection algorithm is O( m 4 κ 2 ).
We present our MFCD algorithm in Algorithm 2, and B. Parameter Determination
facilitate the computational complexity analysis as follows. One important characteristic of our method is that parame-
s s
Matching two cracks Cisi and C j j takes in O(|Cisi ||C j j |) time. ters or thresholds have clear statistical meanings which make
s
Generally, the total number of cracks in I is small and can them easy to be identified for different application scenarios.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Algorithm 2 MFCD Algorithm scales exceeds 3, the detected cracks are almost unchanged.
input : I s , s = 0, 1, . . . , n s − 1 However, the cost of computation increases with this number.
output: C ∗ Therefore, we use 3 scale images. Then, by fixing k = 0, 1, 2,
T p2 log T p 2
we experimentally find σ = 0.8 to be the best choice. Thus,
1 foreach I s do O(n s m4
κ ) the 3 scale images are the original image, σs = 0.8, and
2 Extract C s using WMIP algorithm;
T p2 log T p 2
O( m 4 κ ) σs = 1.6.
s
The choice of m determines the total number of grid cells.
3 foreach C si and C s j do O(n s |Cisi ||C j j |) A smaller m leads to more seeds which increase computation
s s s s
4 foreach Ci i ⊆ C si and C j j ⊆ C s j do O(|Ci i ||C j j |) load in the subsequent steps. On the other hand, increasing
si sj si sj the value of m may miss true crack seeds. In practice,
5 Match Ci and C j ; O(|Ci ||C j |)
m significantly impacts the computation time but its effects
6 foreach C s do O(n s |Cis |) on results do not change much after reaching a sufficient
7 foreach Cis ⊆ C s do O(|Cis |) large number. Thus, we set m = 8 based on the existing
8 Compute ρ(Cis ) using (13); O(|Cis |) literature [34].
Te , Tv and Tr are statistical parameters independent of
9 Merge matched cracks using (14); O(n s κ) image size. They are all based on pixel intensity. Te allows
10 foreach Ci do O(nκ) to retain only the darkest crack seeds. Tv retains the paths
11 Compute f (Ci ) using (15); O(n s ) with the lowest mean intensities. Tr controls the merging
12 if f (Ci ) ≥ F −1 (1 − α, n s ) then of dark pixels neighboring to the currently retained paths.
13 Add Ci into C ∗ ; O(|Ci |) Naturally, we would want maintain Te < Tv < Tr due to
14 return C ∗ ; O(1) their meanings. We choose a dark (i.e. small in intensity)
Te to avoid spending much time on generating many false-
positive paths. A small value of Tv can help us to remove
false-positive paths. However, compared with Te , we prefer a
bigger Tv because there are often pixels which are not as dark
as crack seeds on a path. Meanwhile, a verified path constitutes
only the skeleton of a crack, so a bigger Tr can help to absorb
neighboring dark pixels to grow this skeleton. Thus, we choose
their values jointly as Tr = k1 Tv = K 2 Te , k2 > k1 > 1 to
guarantee Te < Tv < Tr . Tr is determined statistically as
follows: we find the first trough of the image histogram and
count the pixels whose intensities are lower than that of the
first trough, denoting this amount as n, then we determine Tr as
Tr = n/κ, where κ is the total amount of pixels in the image.
For the choice of k1 and k2 , bigger values lead to only discard
few crack seeds and minimal intensity paths which causes
Fig. 5. Pavement inspection robot developed by Guimu Robot Co. Ltd.
significant computation demand in the subsequent steps, while
smaller values increase the risk of removing real cracks. Thus,
Parameters σ p in (11) and α in (17) do not have significant we have to find a trade-off between computing time and
influence on the detection performance. σ p is only used in completeness. In practice, we have found that k1 = 2 and
score computation, and different values only lead to different k2 = 10 work well.
absolute scores of cracks, but still keep the relative differences T p , Tg and Td are self-adaptive parameters which are chosen
between cracks. α is the significance level, indicating the according to the image size. T p determines the maximum
probability of rejecting the null hypothesis when it is true. distance between two adjacent crack seeds to be connected.
A significance level of 0.05 indicates a 5% risk of conclud- If we choose a large T p , we have to spend too much time to
ing that a difference exists when in fact there is no actual generate false-positive paths using Dijkstra algorithm. It does
difference. Thus, in practice, our proposed MFCD algorithm not improve the algorithm results much because most of these
requires the tuning of eight parameters: the standard devia- paths are discarded in the following path verification step.
tion of Gaussian kernel σs for multi-scale image generation, However, if we choose a small T p , some true crack paths
the size of grid cell m, the thresholds Te , Tv , Tr , T p , Tg and Td . may be missed. Considering the concerns above, we choose a
It is important to determine the values of σs in multi- relative bigger value for T p as T p = min(n v , n v )/10, where
scale image generation in Section IV-A. We have to find a n u and n v denote the resolution of the image. Tg and Td are
trade-off between efficiency and completeness. Define σs = only used for the crack matching. A very small value may
kσ, k ∈ N. First, we fix σ = 1, and can adjust the number increases false negative rate, while a very big value may lead
of scales from 1 to 5. By setting k = 0, 1, . . . , n, we obtain to a high false positive rate. In practice, we select at least
n + 1 images with different scales. We select 10 representative 10 representative images for each dataset and we find that
images for the experiments. We find that when the number of Tg = min(n v , n v )/6 and Td = min(n v , n v )/50 work well.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Fig. 6. Computation speed test results of our MFCD algorithm. Each data
point in this figure is an average of the results from ten pavement images.
Fig. 9. Averaged values of Precision, Recall, and F1-measure for all the
images in AigleRN. The values of GC, FFA, MPS are obtained according
to the results provided by [12]. The values of CrackIT and CrackForest are
obtained by running the source codes provided by their respective authors.
Here, WMIP indicates WMIP based crack detection method on I 0 .
Fig. 10. Example results of different algorithms on AiGleRN (from left to right: original image, ground truth, CrackIT, GC, CrackForest, FFA, MPS, WMIP
based method on I 0 , and MFCD).
according to their resolutions: 19 images with a •CrackForest [20]. CrackForest is a road crack detection
resolution of 1200 × 900 pixels, and 14 images with framework based on random structured forests, by learn-
a resolution of 2048 × 2048 pixels. We capture these ing the inherent structured information of cracks.
images with a Genie Nano M1920 Mono camera fixed For FFA, an important parameter is the distance of oriented
on a pavement inspection robot developed by Guimu segments used to compute features. The authors indicate that
Robot Co. Ltd., as shown in Fig. 5. The acquisition are “The distance must be higher than granulate size to obtain
made perpendicular to the pavement, which means that an efficient filtering. If distance is high enough, there is no
the optical axis of the sensor is perpendicular to the noise detection.” In our experiments, we choose the distance
airport runway. The ground truth of cracks are labeled to be 30 when testing on AigleRN and CFD datasets. For
manually by two different experts. Since we have to other methods mentioned above, we use the parameters as
capture these images during the night when the airport is recommended by their authors.
closed, the artificial lighting is used in our system. Due To evaluate the performance of different crack detection
to the special texture, the thin width of cracks and poor algorithms quantitatively, three metrics, including Precision,
lighting conditions, this dataset is quite challenging. Recall and F1-measure, are employed. These three metrics can
The six existing methods which we compare our algorithm be computed based on true positive (TP), false negative (FN),
to are: and false positive (FP),
• GC [35]. GC is a geodesic contour method with automatic TP
selection of points of interest based on auto-correlation. Precision = (20)
TP+FP
• FFA [13]. Free Form Anisotropy (FFA) is based on the TP
estimation of minimal intensity paths at each pixel in four Recall = (21)
TP+FN
directions, and the pixel is recognized as a crack if the Precision × Recall
path cost greatly varies with the direction. F1-measure = 2 × (22)
Precision + Recall
• MPS [12]. Minimal Path Selection (MPS) is a crack
detection algorithm based on the original minimal inten- Since acquiring a high quality ground truth is difficult
sity path selection. for real images, we allow a tolerance margin in measuring
• CrackIT [15]. CrackIT is a Matlab toolbox for road crack the coincidence between the detected cracks and the ground
detection and characterization. truth. As comparisons, according to the experiment settings
• CrackTree [10]. CrackTree is an automatic pavement in [12] and [20] where AigleRN and CFD datasets are
crack detection method which adopts minimum spanning proposed, we assume that TP pixels are included within a
trees to search cracks from a crack seed graph. 2 to 5-pixel vicinity of the ground truth on AigleRN and
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
CFD, respectively. These settings are the same for all tested
methods. For APR dataset, considering that the generation of
the ground truth remains a difficult task in the high resolution
images, we employ a 5-pixel tolerance distance between the
detection and the labeled ground truth for the calculation of
the TP rate.
D. Results on CFD Dataset Fig. 11. Some experimental results using our MFCD algorithm on CFD
dataset (from left to right: original image, ground truth, and results from
Our algorithm also achieves superior performance on CFD MFCD).
dataset. Fig. 11 presents representative images and their detec-
tion results using our MFCD algorithm, where we can see TABLE I
that the images in CFD are quite noisy. The performance of C RACK D ETECTION R ESULTS E VALUATION ON CFD D ATASET
our MFCD algorithm degrades when the contrast between
cracks and backgrounds is low (as shown in the 2nd row
in Fig. 11), or there are dark regions, e.g. oil spots, in the
pavement images (as shown in the 5th and 9th rows in Fig. 11).
The crack detection results on the CFD dataset are summarized
in Tab. I. Again, our algorithm outperforms all counterparts.
E. Results on APR Dataset Fig. 12 presents representative images and their detection
We have tested our algorithm on APR dataset, and com- results using CrackForest, CrackIT and our MFCD algorithms,
pared the results with CrackForest and CrackIT methods. where we can see that the images in APR are very challenging.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
ACKNOWLEDGMENT
The authors would like to thank Guimu Robot Co. Ltd. for
their feedback. They are also grateful to C. Chou, H. Cheng,
S. Yeh, A. Kingery, A. Angert, T. Sun, D. Wang, Y. Sun, Y. Yu,
J. Gong, and M. Momin for their input and contributions to
the Networked Robots Laboratory at Texas A&M University.
R EFERENCES
[1] H. M. La et al., “Mechatronic systems design for an autonomous
robotic system for high-efficiency bridge deck inspection and evalua-
tion,” IEEE/ASME Trans. Mechatronics, vol. 18, no. 6, pp. 1655–1664,
Dec. 2013.
[2] H. M. La et al., “Autonomous robotic system for high-efficiency non-
destructive bridge deck inspection and evaluation,” in Proc. IEEE Int.
Fig. 12. Some experimental results on APR dataset (from left to right: Conf. Automat. Sci. Eng. (CASE), Aug. 2013, pp. 1053–1058.
original image, ground truth, results from CrackForest, CrackIT and MFCD). [3] H. M. La, N. Gucunski, S.-H. Kee, and L. Van Nguyen, “Data analysis
and visualization for the bridge deck inspection and evaluation robotic
system,” Vis. Eng., vol. 3, Dec. 2015, Art. no. 6.
TABLE II [4] Y. Hu, C.-X. Zhao, and H.-N. Wang, “Automatic pavement crack
C RACK D ETECTION R ESULTS E VALUATION ON APR D ATASET detection using texture and shape descriptors,” IETE Tech. Rev., vol. 27,
no. 5, pp. 398–405, 2010.
[5] J. Tang and Y. Gu, “Automatic crack detection and segmentation using
a hybrid algorithm for road distress analysis,” in Proc. IEEE Int. Conf.
Syst., Man, Cybern. (SMC), Oct. 2013, pp. 3026–3030.
[6] Y. Hu and C.-X. Zhao, “A local binary pattern based methods for
pavement crack detection,” J. Pattern Recognit. Res., vol. 5, no. 1,
pp. 140–147, 2010.
[7] R. S. Lim, H. M. La, Z. Shan, and W. Sheng, “Developing a crack
inspection robot for bridge maintenance,” in Proc. IEEE Int. Conf. Robot.
Automat. (ICRA), May 2011, pp. 6288–6293.
The crack detection results on the APR dataset are summarized [8] R. S. Lim, H. M. La, and W. Sheng, “A robotic crack inspection and
in Tab. II where the results of crackIT and CrackForest are mapping system for bridge deck maintenance,” IEEE Trans. Autom. Sci.
Eng., vol. 11, no. 2, pp. 367–378, Feb. 2014.
obtained using their respective open source codes. Note that [9] M. Gavilán et al., “Adaptive road crack detection system by pavement
we did not re-train the model in CrackForest with the images classification,” Sensors, vol. 11, no. 10, pp. 9628–9657, 2011.
of APR dataset. It can be seen that CrackForest produces a [10] Q. Zou, Y. Cao, Q. Li, Q. Mao, and S. Wang, “CrackTree: Automatic
crack detection from pavement images,” Pattern Recognit. Lett., vol. 33,
much higher FP rate. The value of FN rate of CrackIT is a little no. 3, pp. 227–238, 2012.
high that leads to lower recall value. Our MFCD algorithm [11] R. Amhaz, S. Chambon, J. Idier, and V. Baltazart, “A new minimal path
succeeds to detect the cracks in APR dataset and obtain a selection algorithm for automatic crack detection on pavement images,”
in Proc. IEEE Int. Conf. Image Process. (ICIP), Oct. 2014, pp. 788–792.
satisfied overall performance. [12] R. Amhaz, S. Chambon, J. Idier, and V. Baltazart, “Automatic crack
The original images and detection results of the three detection on two-dimensional pavement images: An algorithm based
datasets mentioned above are available on the following web on minimal path selection,” IEEE Trans. Intell. Transp. Syst., vol. 17,
no. 10, pp. 2718–2729, Oct. 2016.
page: http://telerobot.cs.tamu.edu/bridge/Datasets.html. [13] T. S. Nguyen, S. Begot, F. Duculty, and M. Avila, “Free-form anisotropy:
A new method for crack detection on pavement surface images,” in Proc.
VII. C ONCLUSION AND F UTURE W ORK 18th IEEE Int. Conf. Image Process. (ICIP), Sep. 2011, pp. 1069–1072.
[14] H. Oliveira and P. L. Correia, “Automatic road crack detection and
We developed a new MFCD method for detecting the cracks characterization,” IEEE Trans. Intell. Transp. Syst., vol. 14, no. 1,
from pavement images. First, we proposed a WMIP based pp. 155–168, Mar. 2013.
[15] H. Oliveira and P. L. Correia, “CrackIT—An image processing toolbox
method to detect the cracks at individual scales. Second, for crack detection and characterization,” in Proc. IEEE Int. Conf. Image
we matched the corresponding cracks across different scales. Process. (ICIP), Oct. 2014, pp. 798–802.
Finally, a crack evaluation model was built, and the crack was [16] R. Medina, J. Llamas, E. Zalama, and J. Gómez-García-Bermejo,
“Enhanced automatic detection of road surface cracks by combining
selected as the detected crack if it passed statistical hypothesis 2D/3D image processing techniques,” in Proc. IEEE Int. Conf. Image
T 2 log T p
test. Our MFCD algorithm takes in O(n s p m 4 κ 2 ) time. Process. (ICIP), Oct. 2014, pp. 778–782.
[17] M. R. Jahanshahi, S. F. Masri, C. W. Padgett, and G. S. Sukhatme,
We have implemented the proposed algorithm and experi- “An innovative methodology for detection and quantification of cracks
mental results are consistent with our complexity analysis. through incorporation of depth perception,” Mach. Vis. Appl., vol. 24,
We tested MFCD algorithm on three representative datasets. no. 2, pp. 227–241, 2013.
[18] K. Fernandes and L. Ciobanu, “Pavement pathologies classification using
Experimental results showed that 1) The proposed MFCD graph-based features,” in Proc. IEEE Int. Conf. Image Process. (ICIP),
algorithm improves the crack detection performance compared Oct. 2014, pp. 793–797.
with the single scale WMIP based method. 2) Our MFCD [19] P. Prasanna et al., “Automated crack detection on concrete bridges,”
IEEE Trans. Autom. Sci. Eng., vol. 13, no. 2, pp. 591–599, Apr. 2016.
algorithm outperforms six existing methods. [20] Y. Shi, L. Cui, Z. Qi, F. Meng, and Z. Chen, “Automatic road crack
In the future, we will consider fusing the camera inputs detection using random structured forests,” IEEE Trans. Intell. Transp.
with ground penetration radar and laser ranger finder inputs. Syst., vol. 17, no. 12, pp. 3434–3445, Dec. 2016.
[21] A. Cord and S. Chambon, “Automatic road defect detection by tex-
We will study how to perform crack detection using multi- tural pattern recognition based on AdaBoost,” Comput.-Aided Civil
modal inputs. Infrastruct. Eng., vol. 27, no. 4, pp. 244–259, 2012.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
[22] B. Y. Lee, Y. Y. Kim, S.-T. Yi, and J.-K. Kim, “Automated image Dezhen Song (S’02–M’04–SM’09) received the
processing technique for detecting and analysing concrete surface Ph.D. degree in industrial engineering from Univer-
cracks,” Struct. Infrastruct. Eng., vol. 9, no. 6, pp. 567–577, 2013. sity of California at Berkeley, Berkeley, CA, US,
[23] H. M. La, N. Gucunski, K. Dana, and S.-H. Kee, “Development of an in 2004.
autonomous bridge deck inspection robotic system,” J. Field Robot., He is currently a Professor with the Depart-
vol. 34, no. 8, pp. 1489–1504, 2017. ment of Computer Science and Engineering, Texas
[24] D. Zhang, Q. Li, Y. Chen, M. Cao, L. He, and B. Zhang, “An A&M University, College Station, TX, USA. His
efficient and reliable coarse-to-fine approach for asphalt pavement crack research interests include networked robotics, dis-
detection,” Image Vis. Comput., vol. 57, pp. 130–146, Jun. 2017. tributed sensing, computer vision, surveillance, and
[25] S. Chambon and J.-M. Moliard, “Automatic road pavement assessment stochastic modeling.
with image processing: Review and comparison,” Int. J. Geophys., Dr. Song received the Kayamori Best Paper Award
vol. 2011, Jun. 2011, Art. no. 989354. of the 2005 IEEE International Conference on Robotics and Automation (with
[26] V. Kaul, A. Yezzi, and Y. C. Tsai, “Detecting curves with unknown J. Yi and S. Ding). He received the NSF Faculty Early Career Develop-
endpoints and arbitrary topology using minimal paths,” IEEE Trans. ment (CAREER) Award in 2007. From 2008 to 2012, he was an Associate
Pattern Anal. Mach. Intell., vol. 34, no. 10, pp. 1952–1965, Oct. 2012. Editor of the IEEE T RANSACTIONS ON ROBOTICS . From 2010 to 2014,
[27] M. Avila, S. Begot, F. Duculty, and T. S. Nguyen, “2D image based road he was an Associate Editor of IEEE T RANSACTIONS ON AUTOMATION
pavement crack detection by calculating minimal paths and dynamic pro- S CIENCE AND E NGINEERING. He is currently a Senior Editor for IEEE
gramming,” in Proc. IEEE Int. Conf. Image Process. (ICIP), Oct. 2014, ROBOTICS AND AUTOMATION L ETTERS .
pp. 783–787.
[28] Y.-J. Jeon, J. P. Yun, D.-C. Choi, and S. W. Kim, “Defect detection algo-
rithm for corner cracks in steel billet using discrete wavelet transform,”
in Proc. ICCAS-SICE, Aug. 2009, pp. 2769–2773.
[29] R. G. Lins and S. N. Givigi, “Automatic crack detection and measure-
ment based on image analysis,” IEEE Trans. Instrum. Meas., vol. 65,
no. 3, pp. 583–590, Mar. 2016.
[30] L. Ying and E. Salari, “Beamlet transform-based technique for pavement Yu Liu received the B.S. degree from the
crack detection and classification,” Comput.-Aided Civil Infrastruct. Department of Computer Science and Technology,
Eng., vol. 25, no. 8, pp. 572–580, 2010. Taiyuan University of Technology, Taiyuan, China,
[31] R. C. Gonzalez and R. E. Woods, Digital Image Processing, 4th ed. in 2015. He is currently pursuing the master’s
London, U.K.: Pearson, 2017. degree in computer science and technology with the
[32] M. Ester, H.-P. Kriegel, J. Sander, and X. Xu, “A density-based algorithm Civil Aviation University of China, Tianjin, China.
for discovering clusters in large spatial databases with noise,” in Proc. His current research interests include the
KDD, vol. 96. 1996, pp. 226–231. areas of computer vision, image processing, and
[33] E. W. Dijkstra, “A note on two problems in connexion with graphs,” understanding.
Numer. Math., vol. 1, no. 1, pp. 269–271, Dec. 1959.
[34] Y. Huang and B. Xu, “Automatic inspection of pavement cracking
distress,” J. Electron. Imag., vol. 15, no. 1, p. 013017, 2006.
[35] S. Chambon, “Detection of points of interest for geodesic contours:
Application on road images for crack detection,” in Proc. Int. Joint
Conf. Comput. Vis. Theory Appl. (VISAPP), 2011, pp. 1–4.
Haifeng Li received the Ph.D. degree in control the- Binbin Li received the B.S. degree from the
ory and control engineering from Nankai University, Department of Electrical Engineering and Automa-
Tianjin, China, in 2012. tion, Harbin Institute of Technology, Harbin, China,
He is currently an Associate Professor with the in 2012. He is currently pursuing the Ph.D. degree in
Department of Computer Science and Technology, computer engineering with Texas A&M University,
Civil Aviation University of China, Tianjin, China. College Station, TX, USA.
He has authored or co-authored over 30 techni- His current research interests include the areas of
cal articles. His research interests include computer robot vision, visual tracking, and recognition.
vision, image processing, robotic sensing, multisen-
sor fusion, robot localization, and navigation.