CN114596551A - Vehicle-mounted forward-looking image crack detection method - Google Patents

Vehicle-mounted forward-looking image crack detection method Download PDF

Info

Publication number
CN114596551A
CN114596551A CN202210201244.0A CN202210201244A CN114596551A CN 114596551 A CN114596551 A CN 114596551A CN 202210201244 A CN202210201244 A CN 202210201244A CN 114596551 A CN114596551 A CN 114596551A
Authority
CN
China
Prior art keywords
crack
image
formula
gray
cracks
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210201244.0A
Other languages
Chinese (zh)
Inventor
张正鹏
谢欣余
卜丽静
刘科鑫
张嘉裕
戴栋
肖曦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xiangtan University
Original Assignee
Xiangtan University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xiangtan University filed Critical Xiangtan University
Priority to CN202210201244.0A priority Critical patent/CN114596551A/en
Publication of CN114596551A publication Critical patent/CN114596551A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a method for detecting cracks of a vehicle-mounted forward-looking image. The method comprises the steps of road surface semantic segmentation of a forward-looking image, classification of road surface image blocks, crack detection, crack type judgment and statistics. The method has no approximation in the treatment process, can effectively avoid the influence of weed sundries on pavement detection, is clear and convenient for extracting cracks after a histogram gray threshold segmentation algorithm based on significance detection is introduced, can adapt to crack detection of complex images, has general applicability, and can be widely applied to the field of pavement disease detection.

Description

Vehicle-mounted forward-looking image crack detection method
Technical Field
The invention relates to the technical field of road disease detection, in particular to a method for detecting cracks of a vehicle-mounted forward-looking image.
Background
At present, a forward-looking road surface image has the characteristics of diversified background textures, multiple target properties, complex illumination environment, complex acquisition scene and the like, so that a crack becomes one of the most difficult targets to identify in damage types. The vehicle-mounted camera is used for forward-looking shooting, the scene is complex, and the vehicle-mounted camera also comprises auxiliary facilities for road construction and an off-road environment background besides a road; noise complexity, including noise such as shadows, lane lines, oil spots, and water stains;
at present, the pavement crack detection technology mainly comprises three types: artificial visual inspection, digital image processing and three-dimensional laser inspection. The manual visual detection has the problems of low speed, low efficiency, high risk, poor comprehensiveness and the like. The detection step of the image processing-based pavement crack identification technology generally comprises the following steps: the method comprises the steps of firstly collecting crack images through a vehicle-mounted camera, and then preprocessing, crack detection and feature extraction are carried out on the collected images to obtain information such as crack types, crack geometric feature parameters, damage severity and the like (refer to Jiangfeng, Luoku capacity, Zhanghua. concrete road crack detection method based on digital images, review [ J ]. West Hua university newspaper (Nature science edition), 2018(1): 13.). The pavement crack detection technology based on image processing is a common method in the field of pavement crack detection by using the accuracy, safety, robustness and real-time performance of the pavement crack detection technology. The pavement crack detection method based on the three-dimensional laser is free from the interference of illumination and shadow, and has high identification precision and high speed; however, the method has poor detection effect on fine cracks, has high hardware cost and cannot be widely applied.
Therefore, the method has good theoretical significance and practical application value for the research of the vehicle-mounted forward-looking image-based pavement crack detection technology, can provide a more universal, more efficient, more stable and more intelligent pavement crack identification technology for the real-time pavement crack monitoring, and can also provide a scientific evaluation system and a theoretical basis for the pavement crack detection.
Disclosure of Invention
In view of the above technical problems, the present invention is directed to overcome the shortcomings of the prior art, and to provide a method for detecting cracks in a vehicle-mounted forward-view image. The method comprises the steps of road surface semantic segmentation of a forward-looking image, classification of road surface image blocks, crack detection and crack type discrimination and statistics.
In order to achieve the purpose, the invention provides a method for detecting cracks of a vehicle-mounted forward-looking image, which comprises the following steps:
s1: for the vehicle-mounted forward-looking image, separating a road surface and a background area by adopting a CNN road surface semantic segmentation method to obtain a road surface image;
s2: dividing the pavement image obtained in the step S1 into image blocks with n multiplied by n regular sizes, and performing secondary classification on the image blocks by adopting a ResNet network to obtain cracked and non-cracked pavement image blocks;
s3: preprocessing the image blocks of the cracked pavement obtained by the step S2, such as image enhancement, and detecting cracks by adopting an improved HC significance detection and gray histogram threshold segmentation method to obtain a crack segmentation image;
s4: and (4) performing mathematical morphology crack contour extraction on the crack detection image obtained in the step (S3), judging the type of the crack by a histogram projection method, and establishing a two-dimensional coordinate system to calculate the length, the width and other information of the crack.
Preferably, the image preprocessing method in step S3 includes image gray level correction processing, histogram equalization processing, and median filtering denoising processing, the improved HC significance detection method is composed of HC significance detection and fine-scale enhancement, and the crack segmentation is performed by using a gray level histogram threshold segmentation method after the detection, and includes the following contents:
(1) significance detection based on HC: the saliency value of a pixel is defined by the color difference with all other pixels in the image, as follows:
Figure BDA0003529369580000031
in the formula, D (I)K,II) Is a measure of the color distance between two pixels in the space L · a · b, and the above equation, after expanding the pixel levels, can obtain the saliency value of each color as follows:
Figure BDA0003529369580000032
in the formula, c1Is a pixel IkN is the number of different pixel colors, fjIs the pixel color C in the image IjThe frequency of occurrence, true color space, contains 256 by 256 possible colors, more than the total pixels of the image, and is computationally expensive. Each color channel was quantized to 12, the colors were reduced to 12 × 12, the low frequency appearing colors were discarded, and the high frequency appearing colors were retained. The RGB color space quantization method is flawed, and in order to reduce errors, a smoothing operation is used to replace the saliency value of each color with a weighted average of the saliency values of similar colors, but similar colors require distance measurement in the Lab color space. Let m be n/4 nearest neighbor colors to improve color c1The formula is as follows:
Figure BDA0003529369580000041
in the formula (I), the compound is shown in the specification,
Figure BDA0003529369580000042
is the distance between color c and its m nearest neighbors;
(2) the fine scale is obviously enhanced: the fine-scale saliency enhancement algorithm can reduce image noise saliency values. The cracks and the background textures have certain scales, the local features of the cracks can be more highlighted by converting the pavement crack images to the scales suitable for the cracks, and the scale conversion formula is as follows:
It+1(x,y)＝I(UN,t(x,y))×h(·)
where t denotes an image scale, t 0 denotes an original image, and U denotes a pixelN,tAnd (x, y) is the surrounding M multiplied by N field of the pixel point (x, y) under the scale of t, and h (-) is a scale change kernel function. The size of the crack changes due to different image resolutions, the shape of the crack is narrow and long, and the size of the crack should not be too thin. Taking the N as 2, wherein the N is the same as the N,
Figure BDA0003529369580000043
t＝1；
the fine-scale significance enhancement algorithm simulates a linear crack expansion process. The fracture fine-scale significance enhancement formula is as follows:
Figure BDA0003529369580000044
θ∈{0°,45°,90°,135°}
in the formula IuIs the gray scale mean of the crack scale image. Let SzuIs the mean value of significance, wz(x, y) is the corresponding saliency weight at point (x, y), and the formula is as follows:
Figure BDA0003529369580000051
in the formula, GN,θ(x, y) is the sum of the grayscale values of the linear neighborhood in the θ direction centered on the point (x, y), and N is 3. Setting the number of pixels of the original image to be N, wherein the proportion of the candidate points in the original image is at most 10% according to a statistical rule, and when the candidate points are cracks, the linear neighborhood points with the length L around do not need to be screened again;
the gray histogram threshold segmentation method comprises the following steps: for the pavement crack image, when the gray values of the target area and the background area are different, the gray histogram is displayed as two wave crests, and a wave trough is arranged between the wave crests. When the two peaks correspond to the central gray values of the target region and the background region, respectively, the gray value corresponding to the valley value may be used as a threshold for image segmentation. Assuming that the gray value corresponding to the valley is T, then taking T as a segmentation threshold, taking a region composed of pixels with gray values smaller than T as a crack region, and taking pixels with gray values larger than T as a region of a background region. However, the pixel distribution of the crack region and the road surface region follows a normal distribution, that is:
Figure BDA0003529369580000052
Figure BDA0003529369580000053
in the formula (f)1(i) Is a distribution function of the fracture region;
as can be seen from the normal distribution characteristics, a in the above formula1Is the gray scale center value, u, of all pixels in the crack region1Representing the mean grey value, delta, of all pixels within the crack1 2The mean square error of the gray value is represented. Likewise, f2(i) A distribution function representing a background area;
the limiting conditions are as follows:
Figure BDA0003529369580000061
when T lets f (T) take the minimum value, T is the desired threshold. Since the range of gray value variation in the crack image is limited, it does not cover all gray value ranges. Therefore, the method can be used for only reading the pixel number corresponding to the gray value in the crack when designing the algorithm, so that the operation efficiency of the algorithm is improved;
preferably, in the step S4, the fracture contour extraction uses a projection method to identify the fracture image and distinguish the fracture type thereof, and establishes a two-dimensional coordinate system to calculate the length and width information of the fracture, which mainly includes the following methods and features:
(1) crack feature extraction and type identification based on a projection method: and performing feature extraction on the image containing the cracks, and identifying different types of pavement cracks according to different feature values. The method comprises the following steps of carrying out projection in different directions on a binary image after the crack is divided, determining the geometrical characteristics of the crack by counting the number of pixel points, wherein the projection result has the following characteristics:
the horizontal direction projection amplitude of the transverse crack has large difference, and the data change is obvious. When the projection is projected towards the vertical direction, the projection amplitude has obvious difference, but the projection amplitude is smooth. The number of pixel points of each row is approximately the same;
secondly, when the longitudinal crack projects to the horizontal direction, the projection amplitude of the longitudinal crack has obvious difference, and when the longitudinal crack projects to the vertical direction, the projection amplitude of the longitudinal crack has obvious maximum peak value;
the horizontal direction of the reticular cracks is similar to the vertical direction of the projection curve, and the amplitude fluctuation conditions are approximately the same;
(2) quantification of network cracks: the reticular fracture adopts an envelope rectangle to represent the periphery of the fracture, the determination of the peripheral envelope rectangle is judged by searching for the edge points of the fracture target in a fracture segmentation algorithm to obtain four edge points of the envelope rectangle, and then the extreme value of the four edge points is taken as the geometric parameter of the envelope rectangle. Therefore, for the quantification of the network cracks, the damaged area is calculated in an enveloping rectangular manner:
A＝Hμ·Wμ
considering the conversion parameter t of the foresight image coordinate system and the road plane coordinate system, the formula is converted into:
A＝Hμ·Wμ·t
(3) calculation of linear fracture length: because the skeleton of the linear crack is formed by combining a plurality of small sections of skeletons, the length of each small section is calculated and summed, and the obtained result is the length of the crack;
firstly, find out the starting point and the stopping point of the crack skeleton, and assume that the corresponding coordinates are (x) respectively1,y1) And (x)m,yn) Wherein the coordinates of any two adjacent points are (x)t,yt) And (x)t+1,yt+1) The total length of the fracture L can then be calculated according to the following formula:
Figure BDA0003529369580000071
Figure BDA0003529369580000072
in the formula, D is the total length of the image, and Lt is the length of two adjacent points;
considering the conversion parameter t of the forward-looking image coordinate system and the road plane coordinate system, the formula is converted as follows:
Figure BDA0003529369580000073
Figure BDA0003529369580000074
(4) calculation of linear crack width: the crack width can be briefly expressed as the ratio of the sum of the number of pixels in the skeleton to the crack length, and the formula is as follows:
W＝sum/L
considering the conversion parameter t of the forward-looking image coordinate system and the road plane coordinate system, the formula is converted as follows:
W＝sum/L·t
wherein W is the width of the crack;
therefore, the invention provides a vehicle-mounted forward-looking image crack detection method, a deep learning classification method, an HC significance algorithm and an image threshold segmentation algorithm based on a gray histogram are combined for use, the processing process of the method is not approximate, the method can be used in the application fields of road disease detection and the like, and the accuracy and the stability of road crack detection can be effectively improved.
Drawings
The description of the present disclosure will become apparent and readily understood in conjunction with the following drawings, in which:
FIG. 1 is a flow chart of a method for detecting cracks of a vehicle-mounted forward-looking image according to the present invention
FIG. 2 is a diagram of three network authentication results
FIG. 3 image pre-processing map
FIG. 4 comparison graph of significance detection algorithm
FIG. 5 comparison of threshold segmentation methods
FIG. 6 significance detection result binarization
FIG. 7 HC significance detection Algorithm post-thresholding vs. Single histogram-based thresholding
FIG. 8 comparison of skeleton extraction results
Detailed Description
Referring to fig. 1 to 8 and tables 1 to 5, a crack detection method for a vehicle-mounted forward-looking image according to the present invention will be described in detail.
As shown in FIG. 1, the invention relates to a vehicle-mounted foresight image crack detection method, which comprises the following steps:
s1: for the vehicle-mounted forward-looking image, separating a road surface and a background area by adopting a CNN (convolutional neural network) road surface semantic segmentation method to obtain a road surface image;
s2: dividing the pavement image obtained in the step S1 into image blocks with regular size of 10cm multiplied by 10cm, and performing secondary classification on the image blocks by adopting a ResNet network to obtain cracked and non-cracked pavement image blocks;
s3: preprocessing the image blocks of the cracked pavement obtained by the step S2, such as image enhancement, and detecting cracks by adopting an improved HC significance detection and gray histogram threshold segmentation method to obtain a crack segmentation image;
s4: and (4) performing mathematical morphology crack contour extraction on the crack detection image obtained in the step (S3), judging the type of the crack by a histogram projection method, and establishing a two-dimensional coordinate system to calculate the length, the width and other information of the crack.
The process according to the invention is further illustrated by the following working examples.
Data set: according to the collected image data and grid serial numbers, images subjected to chessboard grid blocking are manually classified, the blocked images are classified into two types, namely crack images and non-crack images, all the images with cracks are classified into one type, all the images without cracks are classified into one type, the images are divided into a training set and a verification set, and the training set and the verification set respectively comprise 5000 images and 400 images.
Evaluation indexes are as follows: calculating accuracy (Pr), recall (Re) and F-measure:
a) and (3) separating the road surface and the background area of the acquired image by adopting a CNN road surface semantic segmentation method to obtain a road surface image. And (3) partitioning the segmented road area image into 10cm multiplied by 10cm grids according to the calibration and measurement method of the camera. And (3) performing secondary classification on the image after the grid blocking by adopting a ResNet50 residual network, a VGG network and a MobilenetV2 lightweight mobile network under the same parameters, wherein the parameters are shown in Table 1. The same image block is verified by three methods, namely a VGG convolution neural network, a MobilenetV2 lightweight mobile network and a ResNet50 residual error network, and the obtained results are shown in FIG. 2. The method comprises the steps of selecting 400 pictures, detecting the Top-1 picture accuracy, recall rate and F-measure, and proving that a ResNet50 residual network can effectively classify image blocks with cracks for subsequent crack image preprocessing, detection and identification to meet experiment requirements as shown in Table 2.
TABLE 1
Figure BDA0003529369580000101
TABLE 2
Figure BDA0003529369580000111
b) And preprocessing the classified images with the cracks, and firstly, improving the contrast of the crack regions through gray level conversion to enable the images to reach ideal gray level intervals. Secondly, preprocessing the image through image gray level correction processing, histogram equalization processing and median filtering denoising processing, wherein the processing result is as shown in fig. 3, the noise of the image is removed while the image details are retained and highlighted, and the contradiction between denoising and detail retaining is solved.
c) And after image preprocessing, performing image threshold segmentation of a gray level histogram on the image. The improved HC significance detection algorithm is compared with the FT significance detection algorithm, the AC significance detection algorithm and the LC significance detection algorithm to obtain respective significance maps and carry out binarization, as shown in fig. 4 and 6, and the result proves that the improved HC significance detection algorithm is more suitable for crack detection compared with other three traditional significance detection algorithms.
The iterative global threshold processing, the Otsu global threshold processing and the threshold segmentation method based on the gray histogram are respectively applied to the same image to carry out a comparison experiment, the optimal segmentation method is obtained through analysis, and the experimental result is shown in fig. 6, which proves that the good effect can be obtained by using the valley value of the gray histogram as the threshold to segment the image.
The method combines a gray histogram threshold segmentation algorithm based on an improved HC significance detection algorithm and a threshold segmentation algorithm based on a gray histogram alone to perform segmentation processing on an image and calculate three indexes of accuracy (Precision), Recall (Recall) and F-Measure (F-Measure), the processing result is shown in figure 7, the accuracy, Recall and F-Measure corresponding to a crack detection algorithm are shown in a table 3, and the graph can obviously show that the algorithm of the invention has great improvement on detection and segmentation of pavement cracks, and can effectively segment a target region from a background region.
TABLE 3
Figure BDA0003529369580000121
d) The image after the threshold segmentation processing was subjected to fracture contour extraction based on mathematical morphology, and the result is shown in fig. 8. Through the structured structural elements, the connected domain acquisition and the skeleton extraction are carried out on the image, the contour of the gray level change edge of the binary image can be clearly obtained, the fracture region can be spliced, and isolated points can be effectively removed. And then, extracting and identifying the type of the crack characteristics by adopting a projection method, and distinguishing the crack characteristics into a transverse crack, a longitudinal crack and a reticular crack. And the geometrical characteristic parameters of the cracks are obtained, so that the damage condition of the road surface of the highway can be evaluated accurately, and the actual road surface condition can be reflected.
e) And selecting 40 test chart images for experiment, and comparing the crack extraction result with the crack manual identification result after a series of operations to show the accuracy of the crack identification by the algorithm. After manual detection, 74 transverse cracks, 53 longitudinal cracks and 29 reticular cracks are obtained on the tested section of pavement, and are used as judgment bases, and the crack detection is carried out by utilizing the algorithm used in the text, and partial experimental results are shown in tables 4 and 5.
TABLE 4
Figure BDA0003529369580000131
TABLE 5
Figure BDA0003529369580000132
According to the vehicle-mounted forward-looking image crack detection method, cracks are clear and convenient to extract after a histogram gray threshold segmentation algorithm based on significance detection is introduced, the method can adapt to crack detection of complex images, has general applicability, and can be widely applied to the field of pavement disease detection.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (3)

1. A method for detecting cracks of a vehicle-mounted forward-looking image is characterized by comprising the following steps:
s1: for the vehicle-mounted forward-looking image, separating a road surface and a background area by adopting a CNN road surface semantic segmentation method to obtain a road surface image;
s2: dividing the pavement image obtained in the step S1 into image blocks with n multiplied by n regular sizes, and performing secondary classification on the image blocks by adopting a ResNet network to obtain cracked and non-cracked pavement image blocks;
s3: preprocessing the image blocks of the cracked pavement obtained by the step S2, such as image enhancement, and detecting cracks by adopting an improved HC significance detection and gray histogram threshold segmentation method to obtain a crack segmentation image;
s4: and (4) performing mathematical morphology crack contour extraction on the crack detection image obtained in the step (S3), judging the type of the crack by a histogram projection method, and establishing a two-dimensional coordinate system to calculate the length, the width and other information of the crack.
2. The method for crack detection of vehicle-mounted foresight images as claimed in claim 1, wherein the image preprocessing method in step S3 includes image gray level correction processing, histogram equalization processing and median filtering denoising processing, the improved HC significance detection method is composed of HC significance detection and fine-scale enhancement, and after detection, crack segmentation is performed by using a gray level histogram threshold segmentation method, which mainly includes the following methods and features:
(1) significance detection based on HC: the saliency value of a pixel is defined by the color difference with all other pixels in the image, as follows:
Figure FDA0003529369570000021
in the formula, D (I)K,II) Is a measure of the color distance between two pixels in the space L · a · b, the above formula can obtain a significant value for each color by expanding the pixel levels, and the formula is as follows:
Figure FDA0003529369570000022
in the formula, c1Is a pixel IkN is the number of different pixel colors, fjIs the pixel color C in the image IjThe frequency of occurrence, true color space, contains 256 by 256 possible colors, more than the total pixels of the imageAnd more, the calculation cost is high. Each color channel was quantized to 12, the colors were reduced to 12 × 12, the low frequency appearing colors were discarded, and the high frequency appearing colors were retained. The RGB color space quantization method is flawed, and in order to reduce errors, a smoothing operation is used to replace the saliency value of each color with a weighted average of the saliency values of similar colors, but similar colors require distance measurement in the Lab color space. Let m be n/4 nearest neighbor colors to improve color c1The formula is as follows:
Figure FDA0003529369570000023
in the formula (I), the compound is shown in the specification,
Figure FDA0003529369570000024
is the distance between color c and its m nearest neighbors;
(2) the fine scale is obviously enhanced: the fine-scale saliency enhancement algorithm can reduce image noise saliency values. The cracks and the background textures have certain scales, the local features of the cracks can be more highlighted by converting the pavement crack images to the scales suitable for the cracks, and the scale conversion formula is as follows:
It+1(x,y)＝I(UN,t(x,y))×h(·)
where t denotes an image scale, t 0 denotes an original image, and U denotes a pixelN,tAnd (x, y) is the surrounding M multiplied by N field of the pixel point (x, y) under the scale of t, and h (-) is a scale change kernel function. The crack size is changed due to different image resolutions, the shape of the crack is narrow and narrow, and the size is not excessively thin. Taking the N as 2, wherein the N is the same as the N,
Figure FDA0003529369570000031
t＝1；
the fine-scale significance enhancement algorithm simulates a linear crack expansion process. The fracture fine-scale significance enhancement formula is as follows:
Figure FDA0003529369570000032
θ∈{0°,45°,90°,135°}
in the formula IuIs the gray scale mean of the crack scale image. Let SzuIs the mean value of significance, wz(x, y) is the corresponding saliency weight at point (x, y), and the formula is as follows:
Figure FDA0003529369570000033
in the formula, GN,θ(x, y) is the sum of the grayscale values of the linear neighborhood in the θ direction centered on the point (x, y), and N is 3. Setting the number of pixels of the original image to be N, wherein the proportion of the candidate points in the original image is at most 10% according to a statistical rule, and when the candidate points are cracks, the linear neighborhood points with the length L around do not need to be screened again;
the gray histogram threshold segmentation method comprises the following steps: for the pavement crack image, when the gray values of the target area and the background area are different, the gray histogram is displayed as two wave crests, and a wave trough is arranged between the wave crests. When the two peaks correspond to the central gray values of the target region and the background region, respectively, the gray value corresponding to the valley may be used as a threshold for image segmentation. Assuming that the gray value corresponding to the valley is T, then taking T as a segmentation threshold, taking a region composed of pixels with gray values smaller than T as a crack region, and taking pixels with gray values larger than T as a region of a background region. However, the pixel distribution of the crack region and the road surface region follows a normal distribution, that is:
Figure FDA0003529369570000041
Figure FDA0003529369570000042
in the formula, f1(i) A distribution function for the fracture region;
as can be seen from the normal distribution characteristics, the above formulaIn (a) of1Is the gray scale center value, u, of all pixels in the crack region1Representing the mean grey value, δ, of all pixels within the crack1 2The mean square error of the gray value is represented. Likewise, f2(i) A distribution function representing a background area;
the limiting conditions are as follows:
Figure FDA0003529369570000043
when T lets f (T) take the minimum value, T is the desired threshold. Since the range of gray value variation in the crack image is limited, it does not cover all gray value ranges. Therefore, this can be utilized to read only the number of pixels corresponding to the gray scale value in the crack when designing the algorithm, thereby improving the operation efficiency of the algorithm.
3. The method for crack detection based on vehicle-mounted forward view images as claimed in claim 1, wherein the crack profile extraction in step S4 uses a projection method to identify the crack image and distinguish the crack type, and establishes a two-dimensional coordinate system to calculate the length and width information of the crack, and the method mainly includes the following methods and features:
(1) crack feature extraction and type identification based on a projection method: and performing feature extraction on the image containing the cracks, and identifying different types of pavement cracks according to different feature values. The method comprises the following steps of carrying out projection in different directions on a binary image after the crack is divided, determining the geometrical characteristics of the crack by counting the number of pixel points, wherein the projection result has the following characteristics:
the horizontal direction projection amplitude of the transverse crack has large difference, and the data change is obvious. When the projection is projected towards the vertical direction, the projection amplitude has obvious difference, but the projection amplitude is smooth. The number of pixel points of each row is approximately the same;
secondly, when the longitudinal crack projects to the horizontal direction, the projection amplitude of the longitudinal crack has obvious difference, and when the longitudinal crack projects to the vertical direction, the projection amplitude of the longitudinal crack has obvious maximum peak value;
the horizontal direction of the reticular cracks is similar to the vertical direction of the projection curve, and the amplitude fluctuation conditions are approximately the same;
(2) quantification of network cracks: the reticular fracture adopts an envelope rectangle to represent the periphery of the fracture, the determination of the peripheral envelope rectangle is judged by searching for the edge points of the fracture target in a fracture segmentation algorithm to obtain four edge points of the envelope rectangle, and then the extreme value of the four edge points is taken as the geometric parameter of the envelope rectangle. Therefore, for the quantification of the network cracks, the damaged area is calculated in an enveloping rectangular manner:
A＝Hμ·Wμ
considering the conversion parameter t of the foresight image coordinate system and the road plane coordinate system, the formula is converted into:
A＝Hμ·Wμ·t
(3) calculation of linear fracture length: because the skeleton of the linear crack is formed by combining a plurality of small sections of skeletons, the length of each small section is calculated and summed, and the obtained result is the length of the crack;
firstly, find out the starting point and the stopping point of the crack skeleton, and assume that the corresponding coordinates are (x) respectively1,y1) And (x)m,yn) Wherein the coordinates of any two adjacent points are (x)t,yt) And (x)t+1,yt+1) The total length of the fracture L can then be calculated according to the following formula:
Figure FDA0003529369570000061
Figure FDA0003529369570000062
in the formula, D is the total length of the image, and Lt is the length of two adjacent points;
considering the conversion parameter t of the foresight image coordinate system and the road plane coordinate system, the formula is converted as follows:
Figure FDA0003529369570000063
Figure FDA0003529369570000064
(4) calculation of linear crack width: the crack width can be briefly expressed as the ratio of the sum of the number of pixels in the skeleton to the crack length, and the formula is as follows:
W＝sum/L
considering the conversion parameter t of the foresight image coordinate system and the road plane coordinate system, the formula is converted as follows:
W＝sum/L·t
wherein W is the width of the slit.
CN202210201244.0A 2022-03-03 2022-03-03 Vehicle-mounted forward-looking image crack detection method Pending CN114596551A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210201244.0A CN114596551A (en) 2022-03-03 2022-03-03 Vehicle-mounted forward-looking image crack detection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210201244.0A CN114596551A (en) 2022-03-03 2022-03-03 Vehicle-mounted forward-looking image crack detection method

Publications (1)

Publication Number Publication Date
CN114596551A true CN114596551A (en) 2022-06-07

Family

ID=81816432

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210201244.0A Pending CN114596551A (en) 2022-03-03 2022-03-03 Vehicle-mounted forward-looking image crack detection method

Country Status (1)

Country Link
CN (1) CN114596551A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114882300A (en) * 2022-07-12 2022-08-09 南通翡利达液压科技有限公司 Method and device for identifying significance of scratch in hydraulic bearing super-finishing
CN115311260A (en) * 2022-09-29 2022-11-08 南通羿云智联信息科技有限公司 Road surface quality detection method for highway traffic engineering
CN115830459A (en) * 2023-02-14 2023-03-21 山东省国土空间生态修复中心(山东省地质灾害防治技术指导中心、山东省土地储备中心) Method for detecting damage degree of mountain forest and grass life community based on neural network
CN116448016A (en) * 2023-04-26 2023-07-18 成都智达万应科技有限公司 Intelligent rapid detection system and detection vehicle with same
CN117474912A (en) * 2023-12-27 2024-01-30 浪潮软件科技有限公司 Road section gap analysis method and model based on computer vision

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114882300A (en) * 2022-07-12 2022-08-09 南通翡利达液压科技有限公司 Method and device for identifying significance of scratch in hydraulic bearing super-finishing
CN114882300B (en) * 2022-07-12 2022-09-06 南通翡利达液压科技有限公司 Method and device for identifying significance of scratch in hydraulic bearing super-finishing
CN115311260A (en) * 2022-09-29 2022-11-08 南通羿云智联信息科技有限公司 Road surface quality detection method for highway traffic engineering
CN115311260B (en) * 2022-09-29 2023-12-08 内蒙古公路工程咨询监理有限责任公司 Road surface quality detection method for highway traffic engineering
CN115830459A (en) * 2023-02-14 2023-03-21 山东省国土空间生态修复中心(山东省地质灾害防治技术指导中心、山东省土地储备中心) Method for detecting damage degree of mountain forest and grass life community based on neural network
CN116448016A (en) * 2023-04-26 2023-07-18 成都智达万应科技有限公司 Intelligent rapid detection system and detection vehicle with same
CN117474912A (en) * 2023-12-27 2024-01-30 浪潮软件科技有限公司 Road section gap analysis method and model based on computer vision

Similar Documents

Publication Publication Date Title
CN112419250B (en) Pavement crack digital image extraction, crack repair and crack parameter calculation method
CN114596551A (en) Vehicle-mounted forward-looking image crack detection method
CN107230202B (en) Automatic identification method and system for road surface disease image
CN109087274B (en) Electronic device defect detection method and device based on multi-dimensional fusion and semantic segmentation
CN110443128B (en) Finger vein identification method based on SURF feature point accurate matching
CN113109368B (en) Glass crack detection method, device, equipment and medium
CN107169953B (en) Bridge concrete surface crack detection method based on HOG characteristics
CN101334836B (en) License plate positioning method incorporating color, size and texture characteristic
CN109859226B (en) Detection method of checkerboard corner sub-pixels for graph segmentation
CN107092871B (en) Remote sensing image building detection method based on multiple dimensioned multiple features fusion
CN107330875B (en) Water body surrounding environment change detection method based on forward and reverse heterogeneity of remote sensing image
CN109635733B (en) Parking lot and vehicle target detection method based on visual saliency and queue correction
CN114332026B (en) Visual detection method and device for scratch defects on surface of nameplate
CN107507170A (en) A kind of airfield runway crack detection method based on multi-scale image information fusion
CN111354047B (en) Computer vision-based camera module positioning method and system
CN110070545B (en) Method for automatically extracting urban built-up area by urban texture feature density
CN114863493B (en) Detection method and detection device for low-quality fingerprint image and non-fingerprint image
CN115797813B (en) Water environment pollution detection method based on aerial image
CN111353371A (en) Coastline extraction method based on satellite-borne SAR image
CN115601379A (en) Surface crack accurate detection technology based on digital image processing
CN112101283A (en) Intelligent identification method and system for traffic signs
CN110310263B (en) SAR image residential area detection method based on significance analysis and background prior
Othman et al. Road crack detection using adaptive multi resolution thresholding techniques
CN117635615A (en) Defect detection method and system for realizing punching die based on deep learning
Nguyen et al. A novel automatic concrete surface crack identification using isotropic undecimated wavelet transform

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination