CN116109638B - Rail break detection method and system - Google Patents
Rail break detection method and system Download PDFInfo
- Publication number
- CN116109638B CN116109638B CN202310391551.4A CN202310391551A CN116109638B CN 116109638 B CN116109638 B CN 116109638B CN 202310391551 A CN202310391551 A CN 202310391551A CN 116109638 B CN116109638 B CN 116109638B
- Authority
- CN
- China
- Prior art keywords
- rail
- steel rail
- image
- area
- detection
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 74
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 127
- 239000010959 steel Substances 0.000 claims abstract description 127
- 238000000034 method Methods 0.000 claims abstract description 45
- 238000005070 sampling Methods 0.000 claims abstract description 28
- 238000003384 imaging method Methods 0.000 claims description 28
- 238000012545 processing Methods 0.000 claims description 25
- 239000013598 vector Substances 0.000 claims description 25
- 201000010099 disease Diseases 0.000 claims description 20
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 15
- 238000001914 filtration Methods 0.000 claims description 14
- 230000002159 abnormal effect Effects 0.000 claims description 13
- 230000011218 segmentation Effects 0.000 claims description 7
- 241001270131 Agaricus moelleri Species 0.000 claims description 5
- 238000007637 random forest analysis Methods 0.000 claims description 5
- 238000003909 pattern recognition Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 abstract description 2
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 abstract 2
- 229910052742 iron Inorganic materials 0.000 abstract 1
- 230000007547 defect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241001391944 Commicarpus scandens Species 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
The invention provides a steel rail break detection method and a system, which are used for obtaining an image block by densely sampling a suspected break area which is preliminarily judged, extracting characteristics of the image block, and classifying the image block by adopting a mode identification and AI method, so that the calculation complexity can be effectively reduced. And further utilize texture and depth image, can promote rail break detection accuracy effectively, compare artifical or ultrasonic inspection method, this system has the outstanding advantage that detects fast, can satisfy rail break detection application demand under the track traffic scene such as subway, big iron, high-speed railway.
Description
Technical Field
The invention belongs to the technical field of rail transit disease detection and discloses a method and a system for detecting rail breakage.
Background
The steel rail is an important part of a railway and is used for bearing the weight of a train, and the steel rail is subjected to stress fatigue under the influence of dynamic load in the long-term use process and is easy to break and analyze after long-term use.
The rail breakage will cause major safety accidents, so in the daily railway safety inspection and maintenance process, the rail breakage needs to be detected in time.
The existing rail break detection method is generally an artificial method or ultrasonic flaw detection. However, the manual method has low efficiency, the ultrasonic flaw detection method needs low couplant detection speed and low efficiency, and the long mileage and rapid detection application requirements are difficult to meet.
Disclosure of Invention
In order to solve the problems, the invention provides a steel rail break detection method and a steel rail break detection system, which can realize rapid detection of steel rail breaks.
The technical scheme adopted by the invention is as follows:
a method of detecting rail breakage, the method comprising the steps of: s1, at least acquiring a steel rail texture image and setting a steel rail candidate region R;
s2, judging whether a rail joint area exists in the rail candidate area R, and if the rail joint area exists, shielding the rail joint area to serve as a rail break detection candidate area R1;
s3, in a steel rail break detection candidate region R1, carrying out edge enhancement on the steel rail texture image to obtain an edge enhancement image;
s4, carrying out median filtering on the edge enhanced image to obtain a median filtered image;
s5, searching whether an elongated line exists in the median filtering image;
s6, selecting a line with the length exceeding a set threshold value from the found slender lines;
s7, communicating lines with similar gray scales according to the selected line positions to form a rail break candidate area;
and S8, judging whether the area is a rail break disease according to the rail texture image in the rail break candidate area.
Further, in S2, it is determined whether or not a rail joint region exists in the rail candidate region R by using pattern recognition or AI.
Further, in the step S1, a steel rail depth image may be collected, where the depth image includes shape information of the steel rail;
the method for determining the rail break detection candidate region R1 in the S2 comprises the following steps: in the steel rail depth image, fine positioning is carried out in a steel rail candidate region R to obtain a steel rail detection region RO; judging whether a rail joint area exists in the rail detection area R0, and if the rail joint area exists, shielding the rail joint area to serve as a rail break detection candidate area R1;
also included between S2 and S3 is: in a steel rail break detection candidate region R1, a region lower than the normal height of a steel rail is found out from a steel rail depth image and is used as a first steel rail abnormal region;
s7, communicating lines with similar gray scales according to the selected line positions to serve as a second steel rail abnormal region;
and S8, judging whether the region is a rail break disease according to the rail texture image and the depth image of the rail break candidate region, wherein the rail break candidate region is a region obtained by combining the first rail abnormal region and the second rail abnormal region.
Further, the specific implementation method for judging whether the area is the rail break disease in the S8 is as follows:
s8-1, in a steel rail break candidate area, carrying out dense sampling on the acquired image along the selected line trend to obtain an image block set;
s8-2, classifying the image block sets, and judging whether each image block is broken disease or not;
and S8-3, voting the judgment result of the classification of the image block set, and judging whether the steel rail break candidate area is a break disease or not.
Further, the method for densely sampling the acquired image along the selected line trend is as follows: and carrying out sliding window operation along the trend of the selected line, estimating the main direction of the line in the sampling window, carrying out image sampling after aligning the sampling window with the main direction, and aligning the main direction of the line with the horizontal direction or the vertical direction in a sampled image block.
Further, when the collected image is a steel rail texture image, the method for classifying the image block set comprises the following steps:
carrying out normalization processing on the steel rail texture image, and then carrying out PCA dimension reduction to obtain 1 one-dimensional vector v1;
carrying out transverse and longitudinal accumulated projection on the steel rail texture image to obtain 2 one-dimensional vectors v2 and v3;
the vector V2 and V3 are subjected to normalization processing or normalization and dimension reduction processing and then are connected with the vector V1, so that new feature vectors V0= { V1, V2 and V3} are obtained and used for classification;
training SVM, MLP, KNN or random forest classifier to classify the densely sampled image blocks.
Further, when the collected images are the steel rail texture image and the steel rail depth image, the method for classifying the image block set comprises the following steps:
carrying out normalization processing on the steel rail texture image block and the steel rail depth image block, and then carrying out PCA dimension reduction to obtain 2 one-dimensional vectors v1 and v4;
transversely and longitudinally accumulating and projecting the texture image block and the depth image block to obtain 4 one-dimensional vectors v2, v3, v5 and v6;
the vectors V2, V3, V5 and V6 are subjected to normalization processing or normalization and dimension reduction processing and then are connected with the vectors V1 and V4, so that new feature vectors V1= { V1, V4, V2, V3, V5 and V6} are obtained and used for classification;
training SVM, MLP, KNN or random forest classifier to classify the densely sampled image blocks.
Further, when the acquired images are a steel rail depth image and a steel rail texture image, judging whether a steel rail joint area exists or not by utilizing the steel rail depth image or the steel rail texture image in the steel rail detection area RO;
the method for judging whether the steel rail joint area exists or not by utilizing the steel rail depth image comprises the following steps:
in a steel rail detection area R0, counting a height value h0 of the steel rail;
setting a fishplate detection area by taking a steel rail detection area RO as a reference, setting a height threshold value h1=h0-h 2-t, wherein h2 is the height difference from the top surface of the steel rail to the top surface of the fishplate, and t is an error item;
in the fishplate detection area, a region higher than h1 is found out by a threshold segmentation method;
and selecting a fishplate area from the area, the shape and the depth information serving as constraints in the area higher than h1, and judging that the fishplate area is a rail joint area when the fishplate area exists.
A rail break detection system based on any one of the above detection methods, the system comprising at least:
the image acquisition unit is used for acquiring a steel rail or steel rail texture image and a steel rail depth image;
the image processing unit is used for executing a steel rail break detection algorithm to finish steel rail break detection;
and the carrying unit is used for supplying power to the detection system and installing the support.
Further, a linear array camera and a linear light source are used for acquiring a steel rail texture image, and a linear structured light 3D camera is used for simultaneously acquiring a steel rail depth image and a texture image, wherein the depth image contains shape information of the steel rail.
Further, the imaging resolution of the steel rail texture image along the moving direction is not lower than 0.1mm/pixel, and the imaging resolution along the direction perpendicular to the moving direction is not lower than 0.1mm/pixel;
the imaging resolution of the steel rail depth image and the texture image along the moving direction is not lower than 1mm/pixel, and the imaging resolution along the direction perpendicular to the moving direction is not lower than 1mm/pixel.
The beneficial effects of the invention are as follows:
1. aiming at the rail break detection problem, the rail break detection method based on vision is provided, compared with manual and ultrasonic detection systems, the rail break detection system has the outstanding advantage of high detection speed, can be mounted on high-speed running platforms such as electric buses, and improves detection speed and efficiency.
2. A simple and effective breaking disease judging method is provided: the method is used for obtaining image blocks by densely sampling the suspected broken areas which are primarily judged, extracting features of the image blocks, classifying the image blocks by adopting a mode identification and AI method, so that the computational complexity can be effectively reduced, and the method is convenient to use in low-power consumption application scenes such as detection trolleys and the like.
3. The proposed image dense sampling method comprises the following steps: carrying out sliding window operation along the trend of the line, estimating the main direction of the line in a sampling window, carrying out image sampling after aligning the sampling window with the main direction, and aligning the main direction of the line with the horizontal direction or the vertical direction in a sampled image block; the sampling method reduces the diversity of the rail curve samples and is beneficial to improving the accuracy of image block classification.
4. The line structured light 3D camera is adopted to synchronously acquire the texture image and the depth image of the steel rail with aligned pixels, and the shape of the steel rail is represented by the depth image.
5. The simple and efficient rail joint region detection method is provided: by adopting the depth image and threshold segmentation method, the effective detection of the rail joint is realized, and the actual test shows that the method has high accuracy.
Drawings
FIG. 1 is a schematic diagram of a rail break detection system;
FIG. 2 is a schematic view of a rail break and rail joint;
fig. 3 (a) is a layout diagram of a sampling window of an abnormal region of the steel rail, and fig. 3 (b) is a schematic diagram of alignment of an abscissa of a sampling image block with a main direction;
the device comprises a 1-steel rail, a 2-visual imaging module, a 3-imaging control module, a 4-image processing module, a 5-carrying platform, a 6-breaking area, a 7-steel rail joint, an 8-fishplate, a 9-steel rail joint area, a 10-sampling window and an 11-main direction.
Detailed Description
The present invention will be described in detail below with reference to the drawings and examples thereof, but the scope of the present invention is not limited to the examples.
Example 1: rail break detection method and system when only obtaining rail texture image
The processing steps of the rail break detection method specifically comprise:
s1, acquiring a steel rail texture image by using an image acquisition unit, and setting a steel rail candidate region R in the acquired steel rail texture image;
s2, judging whether a rail joint region 9 exists in the rail candidate region R by adopting a mode identification or AI method (as shown in figure 2); when the rail joint region 9 exists, shielding the rail joint region 9, and then, using the shielded rail joint region as a rail break detection candidate region R1;
s3, in a rail break detection candidate region R1, carrying out high-frequency filtering on a rail texture image by using Fourier transformation, and subtracting the rail image before filtering from the rail image after high-frequency filtering to obtain an edge enhancement image;
s4, median filtering is carried out on the edge enhanced image;
s5, adopting a line filter to find whether an elongated line exists in the median filtering image; if the slender line exists, S6 is carried out, and if the slender line does not exist, ending is carried out;
s6, selecting a line with the length exceeding a set threshold value from the found lines;
s7, communicating lines with similar gray scales according to the selected line positions to form a rail break candidate area;
and S8, judging whether the area is the rail break disease or not by adopting pattern recognition or AI according to the rail texture image in the rail break candidate area.
The specific implementation method of the S8 is as follows:
s8-1, densely sampling the steel rail texture image along the line trend in the steel rail break candidate area to obtain an image block set;
s8-2, classifying the image block sets, and judging whether each image block is broken disease or not;
and S8-3, voting the judgment result of the classification of the image block set, and judging whether the steel rail break candidate area is a break disease or not.
The voting of the judgment result of classifying the image block set in S8-3 is specifically to classify all the image blocks in the image block set in S8-2, judge whether each image block is a breaking disease or a non-breaking disease, and judge whether the steel rail breaking candidate area is a breaking disease according to the majority winning principle. For example, if 10 image blocks are included in S8-2, and 7 image blocks are determined to be broken defects by classification, and 3 image blocks are determined to be non-broken defects, the rail break candidate region is broken defects according to the majority winning principle.
As shown in fig. 3, during sliding window extraction, the main direction of the break area 6 is estimated in the sampling window 10, after aligning the sampling window 10 with the main direction 11, rail texture image sampling is performed, and the main direction 11 in the sampled image block is aligned with the horizontal direction or the vertical direction.
The image block classification method comprises the following steps: carrying out normalization processing on the steel rail texture image blocks, and then carrying out PCA dimension reduction to obtain 1 one-dimensional vector v1; carrying out transverse and longitudinal accumulated projection on the steel rail texture image blocks to obtain 2 one-dimensional vectors V2 and V3, preprocessing the V2 and V3, connecting the preprocessed vectors with a vector V1 to obtain new feature vectors V0= { V1, V2 and V3} for classification, training SVM, MLP, KNN or random forest classifiers, and carrying out dense sampling image block classification. Wherein the pretreatment comprises 2 cases: only normalization processing is carried out, or normalization processing is carried out first and then dimension reduction processing is carried out.
In the rail break detection system of the embodiment, as shown in fig. 1, an image acquisition unit comprises an imaging control module 3 and a visual imaging module 2; the image processing unit is the image processing module 4 in the figure, and the carrying unit is the carrying platform 5 in the figure.
Wherein the imaging control module 3 comprises: the speed measuring unit is used for accurately measuring the speed of the carrying platform; the speed measuring unit is a wheel speed measuring or radar speed measuring or LDV speed measuring module based on an encoder; the imaging control signal generator generates an imaging control pulse signal to the vision imaging module according to the imaging resolution requirements of the movement speed and the movement direction of the operation platform.
The vision imaging module 2 includes: and the linear array camera and the linear light source are used for scanning and imaging the steel rail to obtain a steel rail texture image, wherein the imaging resolution in the moving direction is not lower than 1mm/pixel, and the imaging resolution in the direction perpendicular to the moving direction is not lower than 1mm/pixel.
The image processing module 4 is connected with the imaging control module 3, receives the image acquired by the visual imaging module 2, executes a rail break detection algorithm and completes rail break detection.
The carrying platform 5 is a train or an electric bus or a detection car or a patrol robot or a trolley and provides power supply and installation support for the detection system.
Example 2:
on the basis of embodiment 1, step S8-2 classifies the image block set using a deep learning classification method including, but not limited to, a VGG, resNet, VIT, mobileNet classification model.
Example 3:
the difference from embodiment 1 is that texture images and depth images of the rail surface are obtained simultaneously, at this time, the visual imaging module 2 is a line structured light 3D camera, the imaging resolution of the rail texture images along the moving direction is not lower than 0.1mm/pixel, the imaging resolution along the direction perpendicular to the moving direction is not lower than 0.1mm/pixel, the imaging resolution of the rail depth images along the moving direction is not lower than 1mm/pixel, and the imaging resolution along the direction perpendicular to the moving direction is not lower than 1mm/pixel; the depth image comprises shape information of the steel rail.
The processing steps of the rail break detection method in this embodiment are as follows:
s1, acquiring a steel rail texture image and a steel rail depth image by using an image acquisition unit, and setting a steel rail candidate region R in the acquired steel rail depth image; in the steel rail depth image, a threshold segmentation method is adopted according to the steel rail height information, and a steel rail detection region R0 is finely positioned in a steel rail candidate region R;
s2, judging whether a steel rail joint area exists or not by utilizing a steel rail depth image or a texture image in a steel rail detection area R0; when the rail joint area exists, shielding the rail joint area, and then using the shielded rail joint area as a rail break detection candidate area R1;
s3, in a steel rail break detection candidate region R1, an image threshold segmentation method is adopted, and a region lower than the normal height of the steel rail is found out in a steel rail depth image and is used as a first steel rail abnormal region;
s4, in a rail break detection candidate region R1, carrying out high-frequency filtering on the rail texture image by using Fourier transformation, and subtracting the rail image before filtering from the rail image after high-frequency filtering to obtain an edge enhancement image;
s5, median filtering is carried out on the edge enhanced image;
s6, adopting a line filter to find whether an elongated line exists in the median filtering image;
s7, selecting a line with the length exceeding a set threshold value from the found lines;
s8, communicating lines with similar gray scales according to the selected line positions to serve as a second steel rail abnormal region;
s9, combining the first steel rail abnormal region and the second steel rail abnormal region as steel rail break candidate regions;
and S10, judging whether the area is a rail break disease or not by adopting pattern recognition or AI according to the rail texture image and the depth image in the rail break candidate area.
The specific implementation method of the S10 is as follows:
s10-1, densely sampling texture images and depth images along the trend of a line in a steel rail break candidate area to obtain an image block set;
s10-2, classifying the image block sets, and judging whether each image block is broken disease or not;
and S10-3, judging the steel rail break candidate area as break diseases through voting.
Likewise, the dense sampling method shown in fig. 3 is used for sampling the rail texture image and the depth image. Carrying out normalization processing on the texture image block and the depth image block, and then carrying out PCA dimension reduction to obtain 2 one-dimensional vectors v1 and v4; the texture image block and the depth image block are subjected to transverse and longitudinal accumulated projection to obtain 4 one-dimensional vectors V2, V3, V5 and V6, the vectors V2, V3, V5 and V6 are preprocessed and then connected with the vectors V1 and V4 to obtain new feature vectors V1= { V1, V4, V2, V3, V5 and V6} for classification, and an SVM, MLP, KNN or random forest classifier is trained to classify the densely sampled image blocks. Wherein the pretreatment comprises 2 cases: only normalization processing is carried out, or normalization processing is carried out first and then dimension reduction processing is carried out.
Example 4:
on the basis of embodiment 3, the method for judging whether the rail joint exists or not by using the rail depth image is as follows:
s2-1, counting the height value h0 of the steel rail in a steel rail detection area R0;
s2-2, setting a fishplate detection area R1 by taking RO as a reference, setting a height threshold value h1=h0-h 2-t, wherein h2 is the height difference from the top surface of the steel rail to the top surface of the fishplate, t is an error item, and t is 1-10mm;
s2-3, in R1, finding out a region higher than h1 by a threshold segmentation method;
s2-4, selecting a fishplate area according to the area, shape and depth information as constraint, and judging that the fishplate area is a rail joint when the fishplate area exists; if the fishplate area is not present, it is determined that there is no rail joint area.
Example 5
Based on embodiment 4, the classification method based on deep learning is adopted, including but not limited to VGG, resNet, VIT, mobileNet classification model.
While the specific embodiments of the present invention have been described in detail, the present invention is not limited to the above embodiments, and various changes may be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Claims (6)
1. The method for detecting the breakage of the steel rail is characterized by comprising the following steps of:
s1, acquiring texture images and depth images of a steel rail, setting a steel rail candidate region R, and finely positioning a steel rail detection region R0 in the steel rail candidate region R;
s2, judging whether a rail joint area exists in the rail detection area R0, and if the rail joint area exists, shielding the rail joint area to serve as a rail break detection candidate area R1;
s3, in a steel rail break detection candidate region R1, an image threshold segmentation method is adopted, and a region lower than the normal height of the steel rail is found out in a steel rail depth image and is used as a first steel rail abnormal region;
s4, in a steel rail break detection candidate region R1, carrying out edge enhancement on the steel rail texture image to obtain an edge enhancement image;
s5, median filtering is carried out on the edge enhanced image to obtain a median filtered image;
s6, searching whether an elongated line exists in the median filtering image;
s7, selecting a line with the length exceeding a set threshold value from the found slender lines;
s8, communicating lines with similar gray scales according to the selected line positions to serve as a second steel rail abnormal region;
s9, combining the first steel rail abnormal region and the second steel rail abnormal region as steel rail break candidate regions, and judging whether the regions are steel rail break diseases according to steel rail texture images and depth images in the steel rail break candidate regions;
s9, judging whether the area is the rail break disease or not by the concrete implementation method:
s9-1, in a steel rail break candidate area, carrying out dense sampling on the acquired image along the selected line trend to obtain an image block set;
s9-2, classifying the image block sets, and judging whether each image block is broken disease or not;
s9-3, voting the judgment result of the classification of the image block set, and judging whether the steel rail break candidate area is a break disease or not;
the method for densely sampling the acquired image along the selected line trend comprises the following steps: and carrying out sliding window operation along the trend of the selected line, estimating the main direction of the line in the sampling window, carrying out image sampling after aligning the sampling window with the main direction, and aligning the main direction of the line with the horizontal direction or the vertical direction in a sampled image block.
2. The method for detecting rail breakage according to claim 1, wherein in S2, it is determined whether or not a rail joint region exists in the rail candidate region R by using pattern recognition or AI.
3. The method for detecting the breakage of the steel rail according to claim 1, wherein the method for classifying the image block sets is as follows:
carrying out normalization processing on the steel rail texture image block and the steel rail depth image block, and then carrying out PCA dimension reduction to obtain 2 one-dimensional vectors v1 and v4;
transversely and longitudinally accumulating and projecting the texture image block and the depth image block to obtain 4 one-dimensional vectors v2, v3, v5 and v6;
the vectors V2, V3, V5 and V6 are subjected to normalization processing or normalization and dimension reduction processing and then are connected with the vectors V1 and V4, so that new feature vectors V1= { V1, V4, V2, V3, V5 and V6} are obtained and used for classification;
training SVM, MLP, KNN or random forest classifier to classify the densely sampled image blocks.
4. The method for detecting breakage of a rail according to claim 1, wherein in the rail detection region R0, whether or not a rail joint region exists is determined using a rail depth image and a rail texture image;
the method for judging whether the steel rail joint area exists or not by utilizing the steel rail depth image comprises the following steps:
in a steel rail detection area R0, counting a height value h0 of the steel rail;
setting a fishplate detection area by taking a steel rail detection area R0 as a reference, setting a height threshold value h1=h0-h 2-t, wherein h2 is the height difference from the top surface of the steel rail to the top surface of the fishplate, and t is an error item;
in the fishplate detection area, a region higher than h1 is found out by a threshold segmentation method;
and selecting a fishplate area from the area, the shape and the depth information serving as constraints in the area higher than h1, and judging that the fishplate area is a rail joint area when the fishplate area exists.
5. Rail break detection system based on the detection method according to any of the claims 1-4, characterized in that it comprises at least: the image acquisition unit is used for acquiring a steel rail texture image and a steel rail depth image;
the image processing unit is used for executing a steel rail break detection algorithm to finish steel rail break detection;
and the carrying unit is used for supplying power to the detection system and installing the support.
6. The rail break detection system of claim 5, wherein the imaging resolution of the rail texture image along the movement direction is not less than 0.1mm/pixel, and the imaging resolution along the direction perpendicular to the movement direction is not less than 0.1mm/pixel;
the imaging resolution of the steel rail depth image and the texture image along the moving direction is not lower than 1mm/pixel, and the imaging resolution along the direction perpendicular to the moving direction is not lower than 1mm/pixel.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310391551.4A CN116109638B (en) | 2023-04-13 | 2023-04-13 | Rail break detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310391551.4A CN116109638B (en) | 2023-04-13 | 2023-04-13 | Rail break detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116109638A CN116109638A (en) | 2023-05-12 |
| CN116109638B true CN116109638B (en) | 2023-07-04 |
Family
ID=86258348
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310391551.4A Active CN116109638B (en) | 2023-04-13 | 2023-04-13 | Rail break detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116109638B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102628814A (en) * | 2012-02-28 | 2012-08-08 | 西南交通大学 | Automatic detection method of steel rail light band abnormity based on digital image processing |
| CN105447512A (en) * | 2015-11-13 | 2016-03-30 | 中国科学院自动化研究所 | Coarse-fine optical surface defect detection method and coarse-fine optical surface defect detection device |
| CN107977664A (en) * | 2017-12-08 | 2018-05-01 | 重庆大学 | A kind of road vanishing Point Detection Method method based on single image |
| CN111242891A (en) * | 2019-12-27 | 2020-06-05 | 北京交通大学 | Method for identifying and classifying surface defects of steel rails |
| CN111311567A (en) * | 2020-02-12 | 2020-06-19 | 北京交通大学 | Method for identifying fastener and steel rail diseases of track line image |
| CN113298798A (en) * | 2021-06-10 | 2021-08-24 | 上海电机学院 | Main journal defect detection method based on feature fusion |
| CN114926415A (en) * | 2022-05-06 | 2022-08-19 | 东莞灵虎智能科技有限公司 | Steel rail surface detection method based on PCNN and deep learning |
Family Cites Families (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013020143A1 (en) * | 2011-08-04 | 2013-02-07 | University Of Southern California | Image-based crack quantification |
| KR101360653B1 (en) * | 2012-06-26 | 2014-02-10 | 주식회사 포스코 | Apparatus for detecting seem crack of rolled steel |
| CN104608799B (en) * | 2014-12-12 | 2017-03-29 | 郑州轻工业学院 | Based on information fusion technology Railway wheelset tread damage on-line checking and recognition methodss |
| US10518791B2 (en) * | 2015-10-20 | 2019-12-31 | Sameer Singh | Integrated rail and track condition monitoring system with imaging and inertial sensors |
| CN107576666A (en) * | 2017-10-13 | 2018-01-12 | 成都精工华耀机械制造有限公司 | A kind of double light spectrum image-forming rails and fastener method for detecting abnormality |
| CN108596872B (en) * | 2018-03-08 | 2021-09-28 | 北京交通大学 | Steel rail disease detection method based on Gabor wavelet and SVM |
| WO2020012475A1 (en) * | 2018-07-10 | 2020-01-16 | Rail Vision Ltd | Method and system for railway obstacle detection based on rail segmentation |
| CN110263867A (en) * | 2019-06-24 | 2019-09-20 | 株洲时代电子技术有限公司 | A kind of rail defects and failures classification method |
| CN111626976A (en) * | 2019-11-01 | 2020-09-04 | 东莞灵虎智能科技有限公司 | Steel rail surface damage detection method based on pulse coupling neural network |
| CN111161246B (en) * | 2019-12-30 | 2024-05-14 | 歌尔股份有限公司 | Product defect detection method, device and system |
| CN111444921A (en) * | 2020-03-25 | 2020-07-24 | 浙江华睿科技有限公司 | Scratch defect detection method, device, computing device and storage medium |
| CN112598672A (en) * | 2020-11-02 | 2021-04-02 | 坝道工程医院(平舆) | Pavement disease image segmentation method and system based on deep learning |
| CN112907519A (en) * | 2021-01-29 | 2021-06-04 | 广州信邦智能装备股份有限公司 | Metal curved surface defect analysis system and method based on deep learning |
| CN113903004A (en) * | 2021-10-28 | 2022-01-07 | 北京计算机技术及应用研究所 | Scene recognition method based on middle-layer convolutional neural network multi-dimensional features |
| CN114742768A (en) * | 2022-03-21 | 2022-07-12 | 中车青岛四方车辆研究所有限公司 | Steel rail shallow layer peeling damage detection method and detection system |
| CN114581437A (en) * | 2022-04-09 | 2022-06-03 | 成都精工华耀科技有限公司 | Visual inspection system for visual track guided by visual identification |
-
2023
- 2023-04-13 CN CN202310391551.4A patent/CN116109638B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102628814A (en) * | 2012-02-28 | 2012-08-08 | 西南交通大学 | Automatic detection method of steel rail light band abnormity based on digital image processing |
| CN105447512A (en) * | 2015-11-13 | 2016-03-30 | 中国科学院自动化研究所 | Coarse-fine optical surface defect detection method and coarse-fine optical surface defect detection device |
| CN107977664A (en) * | 2017-12-08 | 2018-05-01 | 重庆大学 | A kind of road vanishing Point Detection Method method based on single image |
| CN111242891A (en) * | 2019-12-27 | 2020-06-05 | 北京交通大学 | Method for identifying and classifying surface defects of steel rails |
| CN111311567A (en) * | 2020-02-12 | 2020-06-19 | 北京交通大学 | Method for identifying fastener and steel rail diseases of track line image |
| CN113298798A (en) * | 2021-06-10 | 2021-08-24 | 上海电机学院 | Main journal defect detection method based on feature fusion |
| CN114926415A (en) * | 2022-05-06 | 2022-08-19 | 东莞灵虎智能科技有限公司 | Steel rail surface detection method based on PCNN and deep learning |
Non-Patent Citations (3)
| Title |
|---|
| 国内外断轨检测技术发展的现状与研究;史宏章等;《铁道运营技术》;第16卷(第4期);1-3,7 * |
| 基于AR的动车组辅助维修系统目标跟踪算法研究;李华玲;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;全文 * |
| 基于超声导波技术的钢轨断裂监测研究;刘建利等;《都市快轨交通》;第35卷(第4期);129-134 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116109638A (en) | 2023-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | A review of applications of visual inspection technology based on image processing in the railway industry | |
| CN115272334B (en) | Method for detecting tiny defects on surface of steel rail under complex background | |
| CN101893580B (en) | Digital image based detection method of surface flaw of steel rail | |
| CN109658397A (en) | A kind of rail polling method and system | |
| CN105260744B (en) | The automatic on-line diagnostic method and system of a kind of goods train coupler yoke key position failure | |
| CN111666947B (en) | Pantograph head offset measuring method and system based on 3D imaging | |
| CN111260629A (en) | Pantograph structure abnormity detection algorithm based on image processing | |
| Karakose et al. | A new approach for condition monitoring and detection of rail components and rail track in railway | |
| CN109489724B (en) | A kind of comprehensive detection device and detection method for safe running environment of tunnel train | |
| CN101706274B (en) | Device for automatically detecting nut loss of rail fastener system | |
| Liu et al. | An efficient method for high-speed railway dropper fault detection based on depthwise separable convolution | |
| KR20190024447A (en) | Real-time line defect detection system | |
| CN109269403A (en) | A kind of contact line conducting wire measurement of wear method and device based on machine vision | |
| CN113487561B (en) | Pantograph foreign matter detection method and device based on gray gradient abnormal voting | |
| CN112419289A (en) | Intelligent detection method for urban subway rail fastener defects | |
| Song et al. | Modeling and optimization of semantic segmentation for track bed foreign object based on attention mechanism | |
| CN112285111A (en) | Pantograph front carbon sliding plate defect detection method, device, system and medium | |
| Li et al. | A visual inspection system for rail corrugation based on local frequency features | |
| CN116109638B (en) | Rail break detection method and system | |
| CN117495759A (en) | Sleeper crack detection method | |
| CN108596968B (en) | Sleeper counting method based on track 3D depth image | |
| CN106482645A (en) | A kind of track ripple grinds detection method | |
| CN111723793B (en) | A real-time rigid catenary positioning point identification method | |
| Anzhong et al. | Multi-target defect detection of railway track based on image processing | |
| CN115641319A (en) | Method and device for detecting current collector of magnetic-levitation train, computer equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |