An image analysis system, method and device suitable for track detection
By collecting and analyzing track images on rail transit vehicles, and calculating curvature by separating line branch method and two-point observation method, the problems of low efficiency and poor accuracy of traditional track detection are solved, and fast and accurate track state detection is achieved, reducing the number of track conditions. Cost and improve security.
Patent Information
- Application Number
- CN202011059164.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-09-30
- Publication Date
- 2025-01-03
- Estimated Expiration
- 2040-09-30
AI Technical Summary
Traditional track detection methods rely on manual or testing vehicles, which are inefficient and costly, making it difficult to achieve rapid and accurate detection of railway tracks, especially under the influence of train movement, which affects the accuracy and safety of the detection results.
Using image analysis method, track images are collected through the image acquisition unit on rail transit vehicles, and the track and vehicle coordinate systems are set using image extraction, positioning, geometric determination and analysis modules, and the curvature is calculated by separating the line and double-point observation method. , judge the track status and realize dynamic displacement detection.
It realizes rapid and accurate detection of railway tracks, reduces detection costs, improves efficiency, eliminates the impact of train motion on detection results, and can perform three-dimensional spatial coordinate positioning, improving the safety and accuracy of detection.
Smart Images
Figure CN114332788B_ABST
Abstract
Description
Technical Field
[0001] The present invention generally relates to an image analysis system, method and apparatus suitable for track inspection. Background Art
[0002] After long-term operation, the position of the railway track will slowly change. When the change exceeds its critical point, the change of the track space position will accelerate. If it is not corrected in time by appropriate detection means, it will not only affect the comfort of passengers, but may also endanger the driving safety of the train in serious cases. However, traditional track inspections mostly rely on manual or track inspection vehicles, and the cost of inspection is relatively high. Moreover, the traditional manual track inspection method is often once every two or three months, which is inefficient and there will be great risks if the inspection is not timely. Therefore, how to quickly detect the geometric status of existing railway lines so as to make necessary repairs to the railway lines in time is the most important issue in railway maintenance.
[0003] Image processing technology provides a track diagnostic tool that can perform more intensive track inspections while trains are in operation, and can screen dangerous sections for further on-site inspection and maintenance. Image processing-based detection methods have developed rapidly in fields such as machinery, and there are currently image processing methods and means for bridge displacement detection. However, existing image processing methods have great limitations on the accuracy of camera lenses and the viewing angle of the object being measured, so it is difficult to apply them to dynamic displacement detection of long-distance tracks. However, train turning, sinking, floating, and nodding movements have a huge impact on image-based track detection, and single-machine images cannot locate the three-dimensional spatial coordinates of the detection track, which greatly affects the accuracy and precision of the track detection results. Summary of the invention
[0004] In view of the above problems in the prior art, the present invention is proposed to solve all or at least one of the above problems.
[0005] According to one aspect of the present invention, there is provided an image analysis method suitable for track detection, the method comprising: an image extraction step, selecting an image including a straight track segment from an image including a track through which a rail transit vehicle passes as an image to start image analysis; an image positioning step, selecting a straight track segment in the selected image, and then setting a track coordinate system and a vehicle coordinate system in the image including the straight track segment, the track coordinate system and the vehicle coordinate system being parallel, wherein the track coordinate system is: the rail corresponding to the straight track segment is the x-axis, the sleeper is the y-axis, and the z-axis is perpendicular to the x-axis and the y-axis; a geometric determination step, in the track coordinate system, selecting an arbitrary point M on the route of the track, the curvature of the point calculated by the secant offset method is called the apparent curvature; and selecting a point N, calculating the curvature at the current moment by the two-point observation method from points M and N; and a track analysis step, judging the state of the track according to the difference between the calculated apparent curvature and the calculated curvature at the current moment.
[0006] According to another aspect of the present invention, there is provided an image analysis system suitable for track detection, the system comprising: an image acquisition unit, which is arranged on a rail transit vehicle to acquire an image of the track when the rail transit vehicle passes through the track; and a processing unit, which processes the acquired image to determine the state of the track, the processing unit comprising: an image extraction module, which is configured to select an image including a straight track segment from an image including the track through which the rail transit vehicle passes as an image to be started for image analysis; an image positioning module, which is configured to select a straight track segment in the selected image, and then locate the straight track segment in the image including the straight track segment; A track coordinate system and a vehicle coordinate system are set in the image, and the track coordinate system and the vehicle coordinate system are parallel, wherein the track coordinate system is: the rail corresponding to the straight track segment is the x-axis, the sleeper is the y-axis, and the z-axis is perpendicular to the x-axis and the y-axis; a geometric judgment module is constructed to select an arbitrary point M on the route of the track in the track coordinate system, and the curvature of the point calculated by the secant offset method is called the apparent curvature; and a point N is selected, and the curvature at the current moment is calculated by the two-point observation method from points M and N; and a track analysis module is constructed to judge the state of the track according to the difference between the calculated apparent curvature and the calculated curvature at the current moment.
[0007] According to another aspect of the present invention, there is provided an image analysis device suitable for track detection, the device comprising: an image extraction module, which is constructed to select an image including a straight track segment from an image including a track through which a rail transit vehicle passes, as an image to start image analysis; an image positioning module, which is constructed to select a straight track segment in the selected image, and then set a track coordinate system and a vehicle coordinate system in the image including the straight track segment, the track coordinate system and the vehicle coordinate system are parallel, wherein the track coordinate system is: the rail corresponding to the straight track segment is the x-axis, the sleeper is the y-axis, and the z-axis is perpendicular to the x-axis and the y-axis; a geometry determination module, which is constructed to select an arbitrary point M on the route of the track in the track coordinate system, and the curvature of the point calculated by the secant offset method is called the apparent curvature; and select a point N, and the curvature at the current moment is calculated by the two-point observation method from points M and N; and a track analysis module, which is constructed to judge the state of the track according to the difference between the calculated apparent curvature and the calculated curvature at the current moment.
[0008] The present invention provides an effective image processing system and method for analyzing images in a train driving recorder to quickly detect the dynamic response of an operating track. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application, and are not limitations to those skilled in the art or ordinary technicians.
[0010] Figure 1 A structural diagram illustrating an example of a track detection image analysis system in the present invention.
[0011] Figure 2 The installation position of the image acquisition unit of the track detection image analysis system in the present invention is illustrated.
[0012] Figure 3 A module diagram of a processing unit in the present invention is illustrated.
[0013] Figure 4 The flowchart of the image analysis method for track detection in the present invention is illustrated.
[0014] Figure 5 The curve and secant offset method for extracting track geometric features in the present invention is illustrated.
[0015] Figure 6 The dual observation method for track state analysis in the present invention is illustrated. DETAILED DESCRIPTION
[0016] Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the following embodiments are not intended to limit the present invention, and that all combinations of the various aspects described in the following embodiments are not necessarily required with respect to the means for solving the problems according to the present invention. For the sake of simplicity, the same marks or numbers are used for the same structural parts or steps, and their descriptions are omitted.
[0017] Track inspection image analysis system
[0018] Attached Figure 1 An example of the track detection image analysis system of the present invention is illustrated. The track detection image analysis system includes: an image acquisition unit, an output unit and a processing unit.
[0019] The image acquisition unit is arranged on a rail transit vehicle to acquire images of the track when the vehicle is traveling. For example, the rail transit vehicle is a train, the track is a train rail, and the image acquisition unit can be a driving recorder 101 arranged on the train.
[0020] The output unit inputs the image captured by the image acquisition unit to the processing unit. An example of the output unit is a video capture card 102, which transmits the image captured by the driving recorder to the processing unit. Optionally, the output unit can be formed integrally with the processing unit or the image acquisition unit, without having to provide an output unit independent of the processing unit / image acquisition unit.
[0021] The processing unit processes the collected images to determine the state of the track. Examples of the driving recorder include a PC 103, a microprocessor, a tablet computer, a cloud processor, etc. In this embodiment, the processing unit is taken as an example of a PC.
[0022] The driving recorder 101 is installed on the train and is used to collect images in front of the train in real time. Figure 2 The images collected by the driving recorder are transmitted to the video capture card 102, and are transmitted to the PC 103 through the video capture card 102 for processing and analysis.
[0023] like Figure 3 As shown, the processing unit 103 may include one or more of the following modules: an image extraction module 310, a calibration module 320, an image positioning module 330, a geometry determination module 340, and a track analysis module 350. The processing performed by each module will be specifically described in the following text in conjunction with the method for track detection image analysis.
[0024] Track Inspection Image Analysis Method
[0025] Examples of the image analysis method of the present invention can be combined with the attached Figure 4 To explain.
[0026] Step S100: collecting video signals of a track when a rail transit vehicle passes through the track, and converting the video signals into image data for storage.
[0027] For example, the driving recorder 101 collects a video signal when a train passes through rails, and converts the video signal into an image for storage.
[0028] Step S120: From the images including the track through which the rail transit vehicle is to pass, select an image including a straight track segment as the image to start image analysis.
[0029] For example, among the images in the video signal collected by the driving recorder 101, an image including a straight track segment is selected for subsequent image analysis.
[0030] Step S140: Select a straight track segment in the selected image, and then set a track coordinate system and a vehicle coordinate system in the image including the straight track segment, wherein the track coordinate system is parallel to the vehicle coordinate system, wherein the track coordinate system is: the rail corresponding to the straight track segment is the x-axis, the sleeper is the y-axis, and the z-axis is perpendicular to the x-axis and the y-axis.
[0031] For example, the plane where the rails of the train are located is selected as the plane formed by the z-axis and the y-axis of the track coordinate system, and the plane where the driving recorder 101 is located in the horizontal direction is selected as the vehicle coordinate system. These two coordinate systems are parallel and separated by a certain distance.
[0032] Optionally, a calibration step may be performed after step S140, by intercepting a straight line segment between two parallel tracks in the track coordinate system and inputting the actual track gauge to calibrate the two coordinate systems.
[0033] Step S160: Geometric determination step, in the track coordinate system, select any point M on the track route, and the curvature of the point calculated by the secant offset method is called the apparent curvature, that is, the dynamic real-time curvature. And select a point N, and calculate the curvature at the current moment by using the double-point observation method from points M and N.
[0034] The following combination Figure 5 The process of geometric determination is described in detail below.
[0035] like Figure 5 As shown, the curve and secant offset method is used to determine the track geometry. When there is no relative motion between the track coordinate system and the vehicle coordinate system, Figure 5 The curve 1 is the direction of the track, the straight line 2 is the direction of the wheel axle, and point B represents the intersection of the track line and the y-axis, and its coordinates are c represents the distance from the intersection of the track and the x-axis to the origin. The following formulas 1 and 2 can be obtained from the Pythagorean theorem:
[0036]
[0037] Where c is the coordinate value of the intersection of the curve and the x-axis on the x-axis. R is the radius of the curve. Subtracting the above formulas yields the following formula 3:
[0038]
[0039] From this, the curvature of the curve can be expressed as the following formula 4:
[0040]
[0041] When there is a relative movement between the vehicle and the track, such as lateral movement, floating, sinking, rolling, shaking and nodding, the rails observed from the carriage will move or rotate accordingly. Figure 6 As shown, the position of the observation point M will be different from the originally observed position, and the rail line shape on the image will also be distorted. The curve seen on the image is called the apparent curve in the present invention. The apparent curve will not only change in position, but also in curvature. That is, the curvature measured by the above formula is the apparent curvature, which is not the actual curvature. Whether it is the apparent curve or the apparent curvature, there is a big problem in how to eliminate the swing of the vehicle in measurement. In order to eliminate this problem, the present invention proposes a dual observation point solution.
[0042] The dual observation point method can eliminate the relative displacement caused by vehicle swing and calculate the consistency of the movement of the vehicle and the track.
[0043] Figure 6 The schematic diagram of double observation is shown, and point N in the figure is another observation point. Compared with the above method, that is,
[0044]
[0045] Similarly, the relationship between the support distance and curvature of the two observation points can be obtained as follows:
[0046]
[0047] When the curvatures obtained by the two methods are equal, there is no relative motion between the vehicle and the track. If they are not equal, there is a relative motion between the vehicle and the track, and the relative motion value is the difference between the two.
[0048] Step S180: Determine the state of the track according to the curvature of the track route in the track coordinate system and the curvature of the track route in the vehicle coordinate system.
[0049] Specifically, the difference between the vehicle coordinate system and the track coordinate system is defined as the wheel-rail fit A, which is expressed by the following formula 6. The greater the wheel-rail fit A, the greater the risk of vehicle derailment:
[0050]
[0051] Wherein, T is the curvature estimation time, i.e., the time required to estimate the curvature. For example, the time corresponding to the vehicle traveling a predetermined length (e.g., the track length is 200 m) is used as the curvature estimation time. W is the standard curvature. In the present invention, When it is a predetermined value (for example, 4 cm), the corresponding curvature of point M is set as the standard curvature. and are the curvatures of the track's course in the track coordinate system and in the vehicle coordinate system, respectively.
[0052] Wheel-rail fit is an indicator that measures the relative motion between the vehicle and the track. If its value is large, it means that the train is at risk of derailment, which is used to judge the state of the track. When the car does not swing relative to the track, the curvature curve characteristic is equal to the eccentricity of the route. Therefore, this method uses the offset value outside the secant line to represent the curvature of the curve.
[0053] The image analysis method of the present invention adopts the curve and secant offset method, with the secant of the arc curve as the reference of the track coordinates, and the two ends of the secant are the positions of the train bogies. The track coordinate system based on the secant is easy to define even if the track has superelevation or the secant is between the curve and the straight line. Since the track is a relative track change based on the vehicle (such as a track inspection vehicle), not a change of the track relative to the geodetic coordinates, the secant coordinate system can clearly illustrate the relative displacement between the vehicle and the track, and can avoid the operation of defining multiple track coordinate systems.
[0054] In the image analysis system of the present invention, analysis errors caused by different image viewing angles can be avoided by defining track, vehicle and viewing angle coordinate systems at the same time.
[0055] The image analysis method of the present invention includes reading images, locating the coordinates of trains and tracks in images, determining track geometry, and analyzing track status. The determination of track geometry includes determining the displacement of specific points or lines on the track, and may also include determining the geometric shape of structures parallel to the driving direction, and determining the changes in image positions and numerical differences between two sets of photos taken at different times.
[0056] In addition, the image analysis of the present invention measures the relative displacement of the vehicle and the track by means of dynamic image measurement, and measures the relative dynamic displacement between the vehicle and the track under the operation of different operating vehicles on the spot, providing the track irregularity along the line. The image analysis software of the present invention can simultaneously perform routine inspections of the track environment, including damage, missing fasteners, water accumulation, etc., as well as analysis and evaluation of vehicle abnormalities or derailment.
[0057] Track inspection image analysis device
[0058] The track detection image analysis device of the present invention comprises: an image extraction module, an image positioning module, a geometry determination module, and a track analysis module. The structures of the above modules are similar to the modules corresponding to the processing unit in the image analysis device, and will not be described in detail here.
[0059] The track detection image analysis according to the present invention can reduce the track detection cost, improve the track detection efficiency, eliminate the huge impact of the train turning, sinking, floating and nodding movements on the image-based track detection, and can use the single-machine image to locate the three-dimensional spatial coordinates of the detection track. Compared with the existing technology, it has outstanding technical effects.
Claims
1. An image analysis method suitable for track detection, the method comprising: An image extraction step, selecting an image including a straight track segment from the image including the track through which the rail transit vehicle passes as the image to be started for image analysis; An image positioning step, selecting a straight track segment in the selected image, and then setting a track coordinate system and a vehicle coordinate system in the image including the straight track segment, wherein the track coordinate system is parallel to the vehicle coordinate system, wherein the track coordinate system is: the rail corresponding to the straight track segment is the x-axis, the sleeper is the y-axis, and the z-axis is perpendicular to the x-axis and the y-axis; The geometric determination step is to select any point M on the track in the track coordinate system, and the curvature of the point calculated by the secant offset method is called the apparent curvature. ; And select another observation point N, and use the double-point observation method to calculate the curvature at the current moment from points M and N ; as well as The orbit analysis step is to calculate the apparent curvature according to the and the calculated curvature at the current moment The difference between and are equal, there is no relative motion between the vehicle and the track, such as and If they are not equal, then there is a relative motion between the vehicle and the track, and the relative motion value is the difference between the two. in, in, is the radius of the track route calculated based on an arbitrary point M on the track, and the coordinates of point M are , point B is an intersection of the track route and the y-axis in the track coordinate system, and its coordinates are , in, is the radius of the path of the track calculated based on an arbitrary point M on the track and the other observation point N, and the coordinates of point M are , the coordinates of point N are .
2. The image analysis method according to claim 1, further comprising: The image acquisition step is to collect the video signal of the track when the rail transit vehicle passes through the track, and convert the video signal into image data storage.
3. The image analysis method according to claim 1, wherein: In the track analysis step, judging the state of the track includes: judging the state of the track according to the wheel-rail tightness A indicating the relative motion between the vehicle and the track, wherein the greater the wheel-rail tightness A is, the greater the risk of vehicle derailment is: Wherein, T is the curvature estimation time, which is the time corresponding to the vehicle traveling a predetermined length, and W is the standard curvature, which is The corresponding curvature of point M when is the predetermined value, and The curvature of the path of the track in the track coordinate system and in the vehicle coordinate system are also represented respectively.
4. The image analysis method according to claim 1, further comprising: The calibration step is to calibrate the two coordinate systems by intercepting a straight line segment between two parallel tracks in the track coordinate system and inputting the actual track gauge.
5. An image analysis system suitable for track detection, the system comprising: An image acquisition unit is provided on a rail transit vehicle to acquire an image of the track when the rail transit vehicle passes through the track; as well as A processing unit processes the collected images to determine the state of the track, the processing unit comprising: An image extraction module is configured to select an image including a straight track segment from an image including the track through which the rail transit vehicle passes as an image to be analyzed; The image positioning module is configured to select a straight track segment in the selected image, and then set a track coordinate system and a vehicle coordinate system in the image including the straight track segment, wherein the track coordinate system is parallel to the vehicle coordinate system, wherein the track coordinate system is: the rail corresponding to the straight track segment is the x-axis, the sleeper is the y-axis, and the z-axis is perpendicular to the x-axis and the y-axis; The geometric determination module is constructed to select an arbitrary point M on the track route in the track coordinate system, and the curvature of the point calculated by the secant offset method is called the apparent curvature ; and select another observation point N, and use the double-point observation method to calculate the curvature at the current moment from points M and N ;as well as A track analysis module is configured to calculate the apparent curvature according to the calculated and the calculated curvature at the current moment The difference between and are equal, there is no relative motion between the vehicle and the track, such as and If they are not equal, then there is a relative motion between the vehicle and the track, and the relative motion value is the difference between the two. in, in, is the radius of the track route calculated based on an arbitrary point M on the track, and the coordinates of point M are , point B is an intersection of the track route and the y-axis in the track coordinate system, and its coordinates are , in, is the radius of the path of the track calculated based on an arbitrary point M on the track and the other observation point N, and the coordinates of point M are , the coordinates of point N are .
6. The image analysis system according to claim 5, wherein: The track analysis module determines the state of the track by: judging the state of the track according to the wheel-rail tightness A indicating the relative motion between the vehicle and the track, wherein the greater the wheel-rail tightness A is, the greater the risk of vehicle derailment is: Wherein, T is the curvature estimation time, which is the time corresponding to the vehicle traveling a predetermined length, and W is the standard curvature, which is The corresponding curvature of point M when is the predetermined value, and The curvature of the path of the track in the track coordinate system and in the vehicle coordinate system are also represented respectively.
7. The image analysis system according to claim 5, further comprising: The calibration module is constructed to calibrate the two coordinate systems by intercepting a straight line segment between two parallel tracks in the track coordinate system and inputting the actual track gauges thereof.
8. The image analysis system according to claim 5, further comprising: The output unit is used to input the image acquired by the image acquisition unit into the processing unit.
9. An image analysis device suitable for track detection, the device comprising: An image extraction module is configured to select an image including a straight track segment from an image including a track through which a rail transit vehicle passes as an image to be analyzed; The image positioning module is configured to select a straight track segment in the selected image, and then set a track coordinate system and a vehicle coordinate system in the image including the straight track segment, wherein the track coordinate system is parallel to the vehicle coordinate system, wherein the track coordinate system is: the rail corresponding to the straight track segment is the x-axis, the sleeper is the y-axis, and the z-axis is perpendicular to the x-axis and the y-axis; The geometric determination module is constructed to select an arbitrary point M on the track route in the track coordinate system, and the curvature of the point calculated by the secant offset method is called the apparent curvature ; And select another observation point N, and use the double-point observation method to calculate the curvature at the current moment from points M and N ; as well as A track analysis module is configured to calculate the apparent curvature according to the calculated and the calculated curvature at the current moment The difference between and are equal, there is no relative motion between the vehicle and the track, such as and If they are not equal, then there is a relative motion between the vehicle and the track, and the relative motion value is the difference between the two. in, in, is the radius of the track route calculated based on an arbitrary point M on the track, and the coordinates of point M are , point B is an intersection of the track route and the y-axis in the track coordinate system, and its coordinates are , in, is the radius of the path of the track calculated based on an arbitrary point M on the track and the other observation point N, and the coordinates of point M are , the coordinates of point N are .