CN114166855A - Real-time rail defect detection method - Google Patents
Real-time rail defect detection method Download PDFInfo
- Publication number
- CN114166855A CN114166855A CN202210129577.7A CN202210129577A CN114166855A CN 114166855 A CN114166855 A CN 114166855A CN 202210129577 A CN202210129577 A CN 202210129577A CN 114166855 A CN114166855 A CN 114166855A
- Authority
- CN
- China
- Prior art keywords
- rail
- defect
- real
- gray
- waveform
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61K—AUXILIARY EQUIPMENT SPECIALLY ADAPTED FOR RAILWAYS, NOT OTHERWISE PROVIDED FOR
- B61K9/00—Railway vehicle profile gauges; Detecting or indicating overheating of components; Apparatus on locomotives or cars to indicate bad track sections; General design of track recording vehicles
- B61K9/08—Measuring installations for surveying permanent way
- B61K9/10—Measuring installations for surveying permanent way for detecting cracks in rails or welds thereof
-
- 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
- G06T7/13—Edge detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/8858—Flaw counting
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/8874—Taking dimensions of defect into account
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Mechanical Engineering (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
A real-time rail defect detection method comprises the following steps: extracting a rail area, determining the left edge and the right edge of a rail, and obtaining a difference image; processing and outputting a rail light band edge oscillogram, searching an optimal threshold value, and outputting a binary defect map in a layering manner; and carrying out defect detection according to the obtained rail light band edge oscillogram and the binary defect map, wherein the defect detection comprises the waveform detection of the rail light band edge oscillogram and the contour detection of the binary defect map, so as to obtain a detection result and a defect type. Compared with the existing defect detection method, the method can realize the detection of various defects, and simultaneously abstract the multi-dimensional characteristics of the defects into one-dimensional wave amplitude, wave crest, quantity and area, thereby reducing the calculation amount.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a real-time rail defect detection method.
Background
With the continuous improvement of the speed, density and load capacity of railway vehicles, rail breakage accidents caused by surface defects of steel rails are also increasing year by year. Therefore, how to detect the surface defects of the steel rails quickly and efficiently is a key problem which needs to be solved for the safety, comfort and high-speed operation of railways. The rail surface defect detection is widely applied to railway safety operation, and the existing rail defect detection methods such as an edge tracking method, a YOLO neural network and the like are started from multiple dimensions, so that larger resources are needed, the real-time performance is not high, and the types of detectable defects are few.
Disclosure of Invention
Based on the problems, the invention provides a real-time rail defect detection method, which reduces multi-dimensional defect characteristics into one-dimensional characteristics after extracting a target by using a dynamic threshold and a difference method, analyzes the wave crest, the amplitude and the frequency of a rail light band edge oscillogram and the defect area and the number of binary defect images to determine the type of the rail defect, thereby being capable of detecting various defect types and improving the detection speed.
A real-time rail defect detection method, comprising the steps of:
step 1: determining the left edge and the right edge of the rail, and extracting a rail area;
step 2: processing and outputting a rail optical band edge oscillogram, obtaining a difference image, searching an optimal threshold value, and outputting a binary defect map in a layering manner;
and step 3: and carrying out defect detection on the obtained rail light band edge oscillogram and the multilayer binary defect map, wherein the defect detection comprises the waveform detection of the rail light band edge oscillogram and the contour detection of the binary defect map, so as to obtain a detection result and a defect type.
Further, the method comprises the steps of:
the step 1 of extracting the rail area comprises the following steps:
step 1-1: converting the input image into a gray level image ImageA;
step 1-2: searching the maximum gray value Max of each line, setting a dynamic threshold value according to the percentage of Max, and searching the coordinates L of the left edge and the right edge of each line of the rail1And R1；
Step 1-3: find all left edge coordinates L1The minimum value being the left edge L of the rail2Find all right edge coordinates R1The central maximum value is the right edge R of the rail2，L2And R2The area between the two is the rail area;
step 1-4: according to the left and right edges L of the rail2And R2And extracting a rail gray level image ImageB.
Further, in the step 2, the light band is a strip-shaped pattern which is brighter than an ungrashed place and is generated on the rail due to the friction between the train and the rail, and the ideal light band is a rectangle, namely, the edge is straight and the internal brightness is uniform.
Further, the step 2 of processing the output rail light band edge waveform map comprises the following steps:
step 2-1: setting a dynamic threshold according to the percentage of the maximum gray value of the rail image, and finding out the left and right edge coordinates L of each row of light band3And R3If more than 3 pixel point rows meeting the threshold are not detected, setting the left and right edge coordinates of the row of light band to be 0 and
；
step 2-2: respectively by the formula L35 and
5 obtaining the change difference L of the left edge and the right edge of the rail light band relative to the left edge and the right edge of the rail4And R4；
Step 2-3: the difference L of the left and right edges of the rail light band relative to the left and right edges of the rail4And R4And the waveform diagram of the edge of the output rail optical band is represented by a broken line diagram.
Further, the step 2 obtains the difference image as follows: in order to enhance the contrast between the defect and the normal rail surface, a formula is adopted
Calculating the gray average value of the ith row pixel point of the rail gray map
Wherein the ith row contains R2-L2+1 pixel points, wherein f (i, j) is the gray value of the corresponding coordinate pixel point; and subtracting the gray average value of the corresponding rail gray image row from the gray value of each row of pixel points of the rail gray image to obtain a difference image ImageC.
Further, the step 2 of finding the optimal threshold specifically includes the following steps:
step 2-4: counting the gray level distribution condition of the difference image ImageC;
step 2-5, two dynamic thresholds are initialized and set to be T respectively according to the gray distribution condition1，T2Dynamic threshold value T1、T2Set according to actual conditions and T1< T2(set to 30% and 50% of maximum gray value, respectively);
step 2-6: a first best threshold search is performed: the gray value of the differential image is less than T1Is greater than T1The pixel point of the image is divided into a foreground part and a background part, and the respective average gray level T of the two parts is calculateda1And Ta2Let T bem1=（Ta1+Ta2) 2 and will Tm1As a new global threshold instead of T1Repeating the above process, and iterating until Tm1Convergence, i.e. Tm1＋1 ＝Tm1Finding the first optimal threshold T3(T for convergence)m1）；
Step 2-7: a second best threshold search is performed: the gray value of the differential image is larger than T1And is less than T2Is greater than T2The pixel point of the image is divided into a foreground part and a background part, and the respective average gray level T of the two parts is calculatedb1And Tb2Let T bem2=（Tb1+Tb2) 2 and will Tm2As a new global threshold instead of T2Repeating the above process, and iterating until Tm2Convergence, i.e. Tm2＋1 ＝Tm2Finding the second optimal threshold T4(T for convergence)m2）。
Further, the step 2 of outputting the binary defect map in a layered manner specifically includes the following steps:
step 2-8: according to the optimum threshold value T4Make the gray value less than T4All the pixel points are set to be 255, and the gray value is larger than T4All the pixel points are set to be 0, and a defect map A is output, wherein the defect map mainly highlights the position of the rail defect;
step 2-9: according to the optimum threshold value T3And T4Make the gray value greater than T3And is less than T4All the pixel points are set to be 255, all the pixel points with other gray values are set to be 0, and a defect map B is output and mainly highlights the scratch defects;
step 2-10: according to the optimum threshold value T3Make the gray value greater than T3All the pixel points are set to be 0, and the gray value is smaller than T3All the pixel points are set to be 255, and a defect map C is output, wherein the defect map mainly highlights pitting corrosion and corrosion.
Further, the step 3 of performing waveform detection on the acquired rail light band edge waveform diagram includes the following steps:
step 3-1: firstly, adopting a sliding average smooth waveform for waveform detection, and then adopting a first derivative to detect the waveform, namely the waveform rises when the first derivative is positive and falls when the first derivative is negative; detecting zero between positive and negative, and then the position is a peak position; if zero is detected between negative and positive, the position is a wave trough; the wave crests are obtained by subtracting the wave troughs from the wave crests, and the wave amplitudes are obtained by subtracting the line numbers corresponding to two adjacent wave troughs. If it is detected that the waveform pattern of the rail band edge is larger than f1The peak of (a) is subjected to the following step, f1The maximum value of the detected wave crest of the normal rail is calculated according to the actual statistics;
step 3-2: respectively detecting the waveform amplitude and the peak value of the waveform diagram of the edge of the rail light band and recording the waveform amplitude and the peak value in arrays Am and Cr;
step 3-3: if there is an amplitude greater than p in Am1If there is no defect, the following steps are carried out, if there is no amplitude larger than p1The waveform of (b) is free from defects, p1Counting the maximum detected amplitude of the normal rail according to actual statistics;
step 3-4: if there is an amplitude greater than p in Am1And is less than p2And its occurrence number of consecutive times is greater than n1Then judge whether the corresponding wave peak is less than f2If less than, there is a saw-tooth defect, n1According to the actual statistics, the continuous times of the corrugation defects are obtained and the minimum value is taken, f2The peak value of the sawtooth defect is obtained according to actual statistics and is taken as the maximumLarge value, p2According to the actual statistics, the amplitude of the corrugation defect is obtained and the minimum value of the amplitude is taken;
step 3-5: if there is an amplitude greater than p in Am2And is less than p3And its occurrence number of consecutive times is greater than n1Then judge whether the corresponding wave crest is larger than f2If it is greater than this, then there is a corrugation defect, p3According to the actual statistics, the amplitude of the abnormal defect is obtained and the minimum value of the amplitude is taken;
step 3-6: if more than p is present in Am3Whether the corresponding peak is larger than f is judged2If the value is larger than the threshold value, an abnormal defect exists, and if the value is smaller than the threshold value, no defect exists.
Further, the step 3 of performing contour detection on the obtained multilayer binary defect map includes the following steps:
step 3-7: carrying out contour detection on the defect image A, wherein the contour detection uses median filtering to filter noise, then adopting a findContours function of Opencv self-carrying to extract contours, if the number of the detected contours is more than 1, carrying out the following steps, otherwise, no defect exists;
step 3-8: and similarly, carrying out contour detection on the defect image B, and if the number of detected contours is more than 1, judging whether the area of the corresponding contour is more than S1If greater than, there is a scratch defect, S1According to the actual statistical scratch area, taking the minimum value;
step 3-9: and similarly, carrying out contour detection on the defect map C, if the number of detected contours is more than n2Judging whether the area of the corresponding contour is smaller than S2If the detected area is less than S, the pit corrosion defect exists, and if the detected area is more than S2Then there is a corrosion defect, n2According to the actual counted number of pitting corrosion and taking the maximum value, S2Is based on actual statistics of the individual erosion areas and takes the minimum.
Compared with the prior art, the invention has the following advantages:
(1) compared with the existing defect detection method, the method can realize multiple defect detections.
(2) Compared with the existing defect detection method, the invention abstracts the multi-dimensional characteristics of the defects into one-dimensional wave amplitude, wave crest, frequency, quantity and area, reduces the calculated amount and improves the detection speed.
Drawings
FIG. 1 is a scene diagram of an example of the real-time rail defect detection method.
FIG. 2 is a flow chart of step 1 in an embodiment of the present invention.
FIG. 3 is a flowchart of step 2 in an embodiment of the present invention.
FIG. 4 is a flowchart illustrating a waveform detection method according to an embodiment of the present invention.
FIG. 5 is a flowchart illustrating binary defect map detection according to an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is further explained in detail by combining the drawings in the specification.
The invention provides a real-time rail defect detection method, which is implemented by the following steps:
firstly reading in image data, then converting the image into a gray scale image, and acquiring the maximum gray scale value Max of each row of pixels of the gray scale image.
Then, defect detection is carried out on the gray-scale image, and the method specifically comprises the following steps:
step 1: as shown in FIG. 2, the dynamic threshold is set as the percentage of the maximum gray value Max of each row, and the coordinates L of the left and right edges of each row of the rail are found1And R1And determining the coordinates of the left edge and the right edge of each row. Find all left edge coordinates L1The minimum value is the left edge L of the rail2Find all right edge coordinates R1The central maximum value is the right edge R of the rail2Determining the left and right edges of the rail as L2And R2，L2And R2And the area between the two areas is the rail area, so that the rail gray level image ImageB is extracted.
Step 2: setting a dynamic threshold according to the percentage of the maximum gray value of the rail image, and finding out the left and right edge coordinates L of each row of light band3And R3，Setting the left and right edge coordinates of the light band of the pixel point row which is not detected to be more than 3 and meets the threshold value as 0 and
. By the formula L35 and
5 obtaining the variation difference L of the left and right edges of the light band relative to the left and right edges of the rail4And R4. Data L4And R4And the waveform diagram of the edge of the output rail optical band is represented by a broken line diagram. In order to enhance the contrast between the defect and the normal rail surface, a formula is adopted
Calculating the gray average value of the ith row pixel point of the rail gray map
Wherein the ith row contains R2-L2+1 pixel points, wherein f (i, j) is the gray value of the corresponding coordinate pixel point; and subtracting the gray average value of the corresponding rail gray image row from the gray value of each row of pixel points of the rail gray image to obtain a difference image ImageC.
As shown in fig. 3, difference image data processing is performed. Firstly, the gray distribution condition of the difference image is counted and a gray histogram is output, and two dynamic thresholds are initialized and set to be T respectively1，T2. Then the best threshold is found: a first best threshold search is performed: the gray value of the differential image is less than T1Is greater than T1The pixel point of the image is divided into a foreground part and a background part, and the respective average gray level T of the two parts is calculateda1And Ta2Let T bem1=（Ta1+Ta2) 2 and will Tm1As a new global threshold instead of T1Repeating the above process, and iterating until Tm1Convergence, i.e. Tm1＋1 ＝Tm1Finding the first optimal threshold T3(ii) a The second best threshold search is performed in the same way: the gray value of the differential image is larger than T1And is less than T2Is greater than T2The pixel point of the image is divided into a foreground part and a background part, and the respective average gray level T of the two parts is calculatedb1And Tb2Let T bem2=（Tb1+Tb2) 2 and will Tm2As a new global threshold instead of T2Repeating the above process, and iterating until Tm2Convergence, i.e. Tm2＋1 ＝Tm2Finding the second optimal threshold T4. And finally, outputting a layered binary defect image: according to the optimum threshold value T4Make the gray value less than T4All the pixel points are set to be 255, and the gray value is larger than T4All the pixel points are set to be 0, and a defect map A is output, wherein the defect map mainly highlights the position of the rail defect; according to the optimum threshold value T3And T4Make the gray value greater than T3And is less than T4All the pixel points are set to be 255, all the pixel points with other gray values are set to be 0, and a defect map B is output and mainly highlights the scratch defects; according to the optimum threshold value T3Make the gray value greater than T3All the pixel points are set to be 0, and the gray value is smaller than T3All the pixel points are set to be 255, and a defect map C is output, wherein the defect map mainly highlights pitting corrosion and corrosion.
And step 3: as shown in fig. 4, waveform detection in defect detection is performed: firstly, adopting a sliding average smooth waveform for waveform detection, and then adopting a first derivative to detect the waveform, namely the waveform rises when the first derivative is positive and falls when the first derivative is negative; detecting zero between positive and negative, and then the position is a peak position; if zero is detected between negative and positive, the position is a wave trough; the wave crests are obtained by subtracting the wave troughs from the wave crests, and the wave amplitudes are obtained by subtracting the line numbers corresponding to two adjacent wave troughs. The waveform is detected if the waveform map of the edge of the rail light band has a peak larger than 0.95 mm. And carrying out waveform amplitude detection and peak value detection on the rail light band edge waveform diagram and recording the waveform amplitude detection and the peak value detection in the arrays Am and Cr respectively. If the Am has a waveform with the amplitude of more than 0.95mm and less than 5.7mm and the occurrence frequency is more than 3, judging whether the corresponding peak is less than 19mm, and if so, determining that the sawtooth defect exists. If the Am has a waveform with the amplitude larger than 5.7mm and smaller than 28.5mm and the occurrence frequency is larger than 3, judging whether the corresponding peak is larger than 19mm, and if so, determining that the corrugation defect exists. If the amplitude of the waveform in Am is larger than 28.5mm, judging whether the corresponding peak is larger than 19mm, and if so, judging that an abnormal defect exists.
As shown in fig. 5, binary map detection in defect detection is performed: and carrying out contour detection on the defect map A, wherein the noise is filtered by adopting median filtering in the contour detection, then the contours are extracted by adopting a findContours function carried by Opencv, and if the number of the contours is detected to be more than 1, carrying out the following steps. Detecting the contour of the defect image B in the same way, and judging whether the area of the corresponding contour is larger than 9.5mm2And if greater, a scratch defect is present. And similarly, carrying out contour detection on the defect map C, and if the number of detected contours is more than 10, judging whether the area of the corresponding contour is less than 3.8mm2If the area is smaller than the area, the pit corrosion defect exists, and if the area is detected to be larger than 3.8mm2Then corrosion defects are present.
An application example of the method is the defect detection of the high-speed rail, and experimental results show that the method can effectively detect various defects and improve certain detection speed.
The above description is only a preferred embodiment of the present invention, and the scope of the present invention is not limited to the above embodiment, but equivalent modifications or changes made by those skilled in the art according to the present disclosure should be included in the scope of the present invention as set forth in the appended claims.
Claims (10)
1. A real-time rail defect detection method, comprising the steps of:
step 1: determining the left edge and the right edge of the rail, and extracting a rail area;
step 2: processing and outputting a rail optical band edge oscillogram, obtaining a difference image, searching an optimal threshold value, and outputting a binary defect map in a layering manner;
and step 3: and carrying out defect detection according to the obtained rail light band edge oscillogram and the multilayer binary defect map, wherein the defect detection comprises the waveform detection of the rail light band edge oscillogram and the contour detection of the binary defect map, so as to obtain a detection result and a defect type.
2. A real-time rail defect detection method according to claim 1, wherein: in the step 1, the left edge and the right edge of the rail are determined according to the dynamic threshold value, and the rail area is extracted.
3. A real-time rail defect detection method according to claim 2, wherein: the dynamic threshold is determined from the percentage of the maximum gray value per row of the gray map.
4. A real-time rail defect detection method according to claim 1, wherein: the step 2 of processing and outputting the waveform diagram of the rail light band edge is as follows: setting a dynamic threshold according to the percentage of the maximum gray value of the rail image, finding the coordinates of the left edge and the right edge of each row of the light band, if more than 3 rows meeting the threshold pixel points are not detected, setting the coordinates of the left edge and the right edge of the row of the light band to be 0 and the width of the rail, and then taking the rail boundary as a reference to obtain the relative change of the left edge and the right edge of the rail light band and representing the relative change by a broken line graph.
5. A real-time rail defect detection method according to claim 1, wherein: the step 2 of obtaining the difference image is: the gray average value of each line of the rail gray image ImageB is firstly obtained, and the gray average value of the corresponding rail gray image line is subtracted from the gray value of each line of pixel points of the rail gray image to obtain a difference image ImageC.
6. A real-time rail defect detection method according to claim 5, wherein: the differential image is intended to enhance the contrast of the defect with the normal rail surface.
7. A real-time rail defect detection method according to claim 1, wherein: the step 2 of finding the optimal threshold value is as follows: initializing and setting two dynamic thresholds according to the condition of analyzing the difference image ImageC gray level distribution, then finding two proper dynamic thresholds by adopting an iterative threshold method, and outputting a binary defect map in a layered mode by utilizing the two found dynamic thresholds to display scratch, pitting and corrosion defects.
8. A real-time rail defect detection method according to claim 1, wherein: and 3, detecting the waveform according to the acquired waveform diagram of the edge of the rail light band as follows: reasonable wave crest, wave amplitude and frequency threshold are set by analyzing the characteristics of three defects of sawtooth, corrugation and light band abnormity; detecting and recording the wave crest and the amplitude of the oscillogram, and judging whether defects exist or not through the wave amplitude; if the defect exists, the size of the wave crest, the size of the wave amplitude and the frequency of the waveform are further analyzed, and the analyzed data are compared with a preset threshold value so as to obtain whether the defect exists and the defect type of the defect.
9. A real-time rail defect detection method according to claim 8, wherein: the waveform detection adopts a sliding average smooth waveform, and then adopts a first derivative to detect the waveform, namely the waveform rises when the first derivative is positive and falls when the first derivative is negative; detecting zero between positive and negative, and then the position is a peak position; if zero is detected between negative and positive, the position is a wave trough; the wave crests are obtained by subtracting the wave troughs from the wave crests, and the wave amplitudes are obtained by subtracting the line numbers corresponding to two adjacent wave troughs.
10. A real-time rail defect detection method according to claim 1, wherein: the step 3 of performing contour detection according to the obtained layered binary defect map comprises the following steps: reasonable area and quantity thresholds are set by analyzing the characteristics of pitting, corrosion and scratching defects; analyzing the outline area and the number of the multilayer binary defect map and comparing the outline area and the number with a preset threshold value to obtain whether defects exist and the defect types of the defects; contour detection filters noise using median filtering and then extracts contours using Opencv's own findContours function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210129577.7A CN114166855B (en) | 2022-02-11 | 2022-02-11 | Real-time rail defect detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210129577.7A CN114166855B (en) | 2022-02-11 | 2022-02-11 | Real-time rail defect detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114166855A true CN114166855A (en) | 2022-03-11 |
| CN114166855B CN114166855B (en) | 2022-04-29 |
Family
ID=80489803
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210129577.7A Active CN114166855B (en) | 2022-02-11 | 2022-02-11 | Real-time rail defect detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114166855B (en) |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH109839A (en) * | 1996-06-24 | 1998-01-16 | Nissan Motor Co Ltd | Surface flaw inspection apparatus |
| 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 |
| US20120263342A1 (en) * | 2011-04-15 | 2012-10-18 | International Business Machines Corporation | Method and system of rail component detection using vision technology |
| CN102854191A (en) * | 2012-07-18 | 2013-01-02 | 湖南大学 | Real-time visual detection and identification method for high speed rail surface defect |
| CN103913464A (en) * | 2014-03-20 | 2014-07-09 | 湖南大学 | High speed railway track surface defect coupling method based on machine visual inspection |
| CN105466941A (en) * | 2015-11-27 | 2016-04-06 | 中国铁道科学研究院 | Steel rail detection method and equipment based on light band image of steel rail |
| CN106683099A (en) * | 2016-11-17 | 2017-05-17 | 南京邮电大学 | Product surface defect detection method |
| CN107194921A (en) * | 2017-05-19 | 2017-09-22 | 华夏高铁技术有限公司 | The automatic identifying method and device of rail rail level light belt |
| WO2018165753A1 (en) * | 2017-03-14 | 2018-09-20 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN109878599A (en) * | 2019-03-22 | 2019-06-14 | 路仁电子科技河北有限公司 | A kind of small rail car device for track multivariate data synchronous acquisition and detection |
| CN112419289A (en) * | 2020-11-30 | 2021-02-26 | 西南交通大学 | Intelligent detection method for urban subway rail fastener defects |
| CN112767326A (en) * | 2021-01-07 | 2021-05-07 | 湖南大学 | Steel rail surface defect detection method, device, equipment and storage medium |
| CN113362285A (en) * | 2021-05-21 | 2021-09-07 | 同济大学 | Steel rail surface damage fine-grained image classification and detection method |
| CN113379728A (en) * | 2021-07-02 | 2021-09-10 | 上海电气集团股份有限公司 | Method, system, equipment and readable storage medium for detecting defects on surface of rail |
| CN214794771U (en) * | 2021-05-20 | 2021-11-19 | 合肥超科电子有限公司 | Contact type red light band generating device for steel rail flaw detection vehicle |
-
2022
- 2022-02-11 CN CN202210129577.7A patent/CN114166855B/en active Active
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH109839A (en) * | 1996-06-24 | 1998-01-16 | Nissan Motor Co Ltd | Surface flaw inspection apparatus |
| US20120263342A1 (en) * | 2011-04-15 | 2012-10-18 | International Business Machines Corporation | Method and system of rail component detection using vision technology |
| EP2697738A1 (en) * | 2011-04-15 | 2014-02-19 | International Business Machines Corporation | Method and system of rail component detection using vision technology |
| 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 |
| CN102854191A (en) * | 2012-07-18 | 2013-01-02 | 湖南大学 | Real-time visual detection and identification method for high speed rail surface defect |
| CN103913464A (en) * | 2014-03-20 | 2014-07-09 | 湖南大学 | High speed railway track surface defect coupling method based on machine visual inspection |
| CN105466941A (en) * | 2015-11-27 | 2016-04-06 | 中国铁道科学研究院 | Steel rail detection method and equipment based on light band image of steel rail |
| CN106683099A (en) * | 2016-11-17 | 2017-05-17 | 南京邮电大学 | Product surface defect detection method |
| WO2018165753A1 (en) * | 2017-03-14 | 2018-09-20 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN107194921A (en) * | 2017-05-19 | 2017-09-22 | 华夏高铁技术有限公司 | The automatic identifying method and device of rail rail level light belt |
| CN109878599A (en) * | 2019-03-22 | 2019-06-14 | 路仁电子科技河北有限公司 | A kind of small rail car device for track multivariate data synchronous acquisition and detection |
| CN112419289A (en) * | 2020-11-30 | 2021-02-26 | 西南交通大学 | Intelligent detection method for urban subway rail fastener defects |
| CN112767326A (en) * | 2021-01-07 | 2021-05-07 | 湖南大学 | Steel rail surface defect detection method, device, equipment and storage medium |
| CN214794771U (en) * | 2021-05-20 | 2021-11-19 | 合肥超科电子有限公司 | Contact type red light band generating device for steel rail flaw detection vehicle |
| CN113362285A (en) * | 2021-05-21 | 2021-09-07 | 同济大学 | Steel rail surface damage fine-grained image classification and detection method |
| CN113379728A (en) * | 2021-07-02 | 2021-09-10 | 上海电气集团股份有限公司 | Method, system, equipment and readable storage medium for detecting defects on surface of rail |
Non-Patent Citations (19)
| Title |
|---|
| Q. D. Y. MA 等: "Artificial object edge detection based on enhanced Canny algorithm for high-speed railway apparatus identification", 《2017 10TH INTERNATIONAL CONGRESS ON IMAGE AND SIGNAL PROCESSING》 * |
| 乔湘洋等: "基于机器视觉的线缆表面缺陷检测系统设计与算法研究", 《机床与液压》 * |
| 冯凯: "基于图像的轨道缺陷自动检测系统的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 吴成和 等: "基于数字图像处理的轨道板裂纹缺陷检测", 《光电技术应用》 * |
| 唐湘娜等: "铁轨表面缺陷的视觉检测与识别算法", 《计算机工程》 * |
| 康高强等: "基于激光摄像技术的钢轨波磨检测方法研究", 《城市轨道交通研究》 * |
| 张辉 等: "钢轨缺陷无损检测与评估技术综述", 《仪器仪表学报》 * |
| 曹义亲等: "基于缺陷比例限制的背景差分钢轨表面缺陷检测方法", 《计算机应用》 * |
| 李唯一等: "铁路钢轨光带异常检测系统", 《计算机工程与应用》 * |
| 李锦珑等: "基于机器视觉的铁轨扣件边缘检测研究", 《西北师范大学学报(自然科学版)》 * |
| 武丹: "基于图像的轨道表面缺陷检测及分类算法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
| 王昊: "高速轨道检测图像处理技术", 《铁路技术创新》 * |
| 甄理: "基于计算机视觉的铁轨表面缺陷检测技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 米曾真等: "重轨图像增强与边缘提取的关键技术", 《光学精密工程》 * |
| 茅正冲 等: "基于图像传感器的铁轨表面缺陷视觉检测算法", 《传感器与微系统》 * |
| 许玉德等: "手推式钢轨磨耗状态检测仪的研制", 《城市轨道交通研究》 * |
| 谭云强: "基于计算机视觉的钢轨缺陷算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 谭松等: "基于计算机视觉的车载钢轨光带异常检测系统研制", 《铁道建筑》 * |
| 钟瑞: "基于计算机视觉的钢轨光带几何特征检测", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114166855B (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115170576B (en) | Aluminum pipe surface defect detection method based on machine vision | |
| CN115457041B (en) | Road quality identification and detection method | |
| CN107230202B (en) | Automatic identification method and system for road surface disease image | |
| CN106251361B (en) | A kind of rail surface defects image self-adapting division method | |
| CN108647572B (en) | Lane departure early warning method based on Hough transform | |
| CN115049664B (en) | Vision-based ship engine fitting defect detection method | |
| CN114937055A (en) | Image self-adaptive segmentation method and system based on artificial intelligence | |
| CN115330783A (en) | Steel wire rope defect detection method | |
| Qingbo | Pavement crack detection algorithm based on image processing analysis | |
| Liang et al. | Defect detection of rail surface with deep convolutional neural networks | |
| CN114972326A (en) | Defective product identification method for heat-shrinkable tube expanding process | |
| CN103455994A (en) | Method and equipment for determining image blurriness | |
| CN107944354B (en) | Vehicle detection method based on deep learning | |
| CN116385450B (en) | PS sheet wear resistance detection method based on image processing | |
| CN115496692B (en) | Lubricating oil abrasive particle image enhancement method | |
| CN107464256A (en) | A kind of target detection differentiates the correlating method of amendment with possibility | |
| CN114820625B (en) | Automobile top block defect detection method | |
| CN111738256A (en) | Composite material CT image segmentation method based on improved watershed algorithm | |
| CN109850518B (en) | Real-time mining adhesive tape early warning tearing detection method based on infrared image | |
| CN111476804A (en) | Method, device and equipment for efficiently segmenting carrier roller image and storage medium | |
| CN114166855B (en) | Real-time rail defect detection method | |
| CN102509265B (en) | Digital image denoising method based on gray value difference and local energy | |
| CN117496109A (en) | Image comparison and analysis method and device, electronic equipment and storage medium | |
| CN113516608A (en) | Tire defect detection method and device, and tire detection device | |
| CN113834447B (en) | High-dynamic laser light bar self-adaptive imaging processing method in outdoor complex environment |
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 |