A method for measuring the swing angle of heavy-duty train couplers based on machine vision
Through a machine vision-based method, deep learning models are used to identify and calculate the swing angle of the heavy-loaded train hook, which solves the problems of installation difficulties, easy sensor damage and high failure rate in the prior art, and realizes high-precision, low-cost and low-hidden danger hook swing angle monitoring.
Patent Information
- Application Number
- CN202510290675.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-12
- Publication Date
- 2025-05-06
- Estimated Expiration
- 2045-03-12
AI Technical Summary
The existing method of coupling swing angle monitoring of heavy-load trains has problems such as difficulty in installing equipment, easy sensor damage and high failure rate, making it difficult to achieve high-precision long-term and stable monitoring.
Using a machine vision-based method, the hook area image is collected through the camera, preprocessing and the hook feature target recognition is performed, and the hook swing angle is calculated using deep learning models such as the improved yolov9 network model.
It realizes high-precision non-contact measurement of the swing angle of the heavy-duty train, reduces installation and maintenance costs, reduces safety hazards, and is suitable for long-term monitoring.
Smart Images
Figure CN119810099B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of heavy-load train coupler swing angle recognition, and in particular to a heavy-load train coupler swing angle measurement method based on machine vision. Background Art
[0002] Heavy-haul trains are known for their long formations and large axle loads, and play a vital role in improving the efficiency of heavy-haul railway transportation. China's heavy-haul railways are characterized by long slopes and many curves. On long downhill sections, when the train speed is controlled by the circulating air braking strategy, the train will experience a strong longitudinal inertial impulse due to the asynchronous braking and relief of different vehicles. As an important component of heavy-haul trains, the coupler may lose stability after being subjected to longitudinal impact, resulting in a large coupler swing angle. As the coupler swing angle continues to increase, the lateral component of the coupler force will increase, thereby increasing the wheel-rail lateral force and the locomotive derailment coefficient. Once the above two safety indicators exceed the threshold, the operational safety of the train cannot be guaranteed. Therefore, real-time monitoring of the coupler swing angle is crucial to ensure the operational safety of heavy-haul trains.
[0003] At present, researchers have proposed several feasible on-board coupler swing angle monitoring schemes, which can be mainly divided into direct measurement methods and indirect measurement methods. The traditional direct measurement method is currently widely used. This method is based on the geometric relationship between several points on the coupler and calculates the coupler swing angle through the dynamic displacement change of the displacement sensor. This method is simple in principle and has high measurement accuracy. However, there are obvious disadvantages in installing a large number of measuring equipment at the coupler reconnection point. On the one hand, this method requires the installation of a large number of instruments and tooling in the limited space of the coupler reconnection point, which may interfere with other vehicle components. On the other hand, due to the harsh service environment of the equipment, the sensors installed on the coupler and their fixed tooling may be damaged or fall off due to severe vibration. This will not only affect the stability of the coupler swing angle measurement, but may even seriously threaten the safety of train operation. To improve these problems, some researchers have proposed an indirect coupler swing angle measurement method to quantitatively identify the coupler swing angle based on the relative displacement of the locomotive body frame. However, the existing indirect measurement method essentially still relies on the installation of a large number of contact displacement sensors to identify the coupler swing angle, and still has the disadvantages of high failure rate and greater safety hazards. Therefore, developing a non-contact coupler swing angle measurement method based on machine vision will effectively overcome the limitations of existing methods and realize high-precision, long-term and stable monitoring of the coupler swing angle, which is of great significance for ensuring the safe operation of heavy-load trains. Summary of the invention
[0004] In view of the above problems, the present invention aims to provide a method for measuring the swing angle of a heavy-load train coupler based on machine vision.
[0005] The technical solution of the present invention is as follows:
[0006] A method for measuring the swing angle of a heavy-load train coupler based on machine vision comprises the following steps:
[0007] S1: collecting an image of a coupler area of a heavy-load train, and preprocessing the image to obtain a preprocessed coupler area image;
[0008] S2: performing coupler feature target recognition on the preprocessed coupler region image to obtain pixel coordinates of the coupler;
[0009] S3: Calculate the coupler swing angle according to the pixel coordinates of the coupler.
[0010] Preferably, in step S1, the image of the coupler area of the heavy-load train is obtained by a camera installed on the vehicle body just above the coupler.
[0011] Preferably, in step S1, the preprocessing includes image enhancement processing.
[0012] Preferably, the image enhancement processing includes adding Gaussian noise, adjusting brightness and random exposure processing.
[0013] Preferably, in step S2, a deep learning model is used to perform coupler feature target recognition on the preprocessed coupler area image.
[0014] Preferably, the deep learning model adopts any one of the Yolo series models.
[0015] Preferably, the deep learning model adopts an improved yolov9 network model, which, based on the yolov9 network model, adds a SEAM attention mechanism before its six DETECT output modules, and adds a ByteTrack tracking algorithm module after the three DETECT output modules that are always enabled during its model training and recognition.
[0016] Preferably, in step S3, the coupler swing angle is calculated by the following formula:
[0017] (1)
[0018] Where: is the hook swing angle; is the number of pixels corresponding to the unit physical size on the horizontal pixel axis; is the number of pixels corresponding to the unit physical size on the pixel vertical axis; is the horizontal coordinate of the pixel coordinate of the coupler; It is half of the total length of the horizontal axis of the image pixels; It is the difference in pixel horizontal coordinates from the center point of the round pin to the center axis of the coupler; The actual longitudinal distance between the camera and the connection point between the coupler and the vehicle body; is the pitch angle of the camera; is the center point of the pin in the world coordinate system Coordinates in the axis direction; is the ordinate of the pixel coordinate of the coupler; It is half of the total length of the vertical axis of the image pixels; is the focal length of the camera.
[0019] Preferably, in step S3, the coupler swing angle is calculated by the following formula:
[0020] (2)
[0021] Where: is the hook swing angle; , They are the horizontal and vertical coordinates of the pixel coordinates of the coupler respectively; , , All are calibration constants.
[0022] Preferably, the calibration constant is obtained by calibration through the following steps: obtaining the measured coupler swing angle and coupler pixel coordinates, using formula (2) to establish equations about the calibration constant under at least three groups of different coupler swing angle conditions; solving the equation to obtain the calibration constant.
[0023] The beneficial effects of the present invention are:
[0024] The present invention quantitatively identifies the coupler swing angle of heavy-load trains based on machine vision, thereby realizing high-precision non-contact measurement of the coupler swing angle of heavy-load trains. It has the characteristics of low installation and maintenance costs and small safety hazards, and is expected to develop into a long-term monitoring method for the coupler swing angle of heavy-load trains. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0026] Figure 1 Schematic diagram of the installation position of the camera of the present invention;
[0027] Figure 2 This is a schematic diagram of the structure of the improved yolov9 network model of the present invention;
[0028] Figure 3Schematic diagram of coordinate transformation of pixel coordinates of a coupler according to the present invention;
[0029] Figure 4 This is a geometric diagram of the calculation of the coupler swing angle of the present invention;
[0030] Figure 5 It is a schematic diagram of the original image and preprocessing result of the coupler area in a specific embodiment;
[0031] Figure 6 It is a schematic diagram of visualization results of coupler feature target recognition in a specific embodiment;
[0032] Figure 7 It is a schematic diagram of the measurement result of the swing angle of the coupler under the condition of small swing angle in a specific embodiment;
[0033] Figure 8 It is a schematic diagram of the measurement results of the coupler swing angle under large swing angle conditions in a specific embodiment. DETAILED DESCRIPTION
[0034] The present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, in the absence of conflict, the embodiments in this application and the technical features in the embodiments can be combined with each other. It should be noted that, unless otherwise specified, all technical and scientific terms used in this application have the same meanings as those generally understood by those of ordinary skill in the art to which this application belongs. The words "including" or "comprising" and the like used in the disclosure of the present invention mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects.
[0035] The present invention provides a method for measuring the swing angle of a heavy-duty train coupler based on machine vision, comprising the following steps:
[0036] S1: collecting an image of a coupler region of a heavy-load train, and preprocessing the image to obtain a preprocessed coupler region image.
[0037] In a specific embodiment, Figure 1 As shown, the image of the heavy-load train coupler area is obtained by a camera installed on the car body directly above the coupler. It should be noted that the image obtained directly above is more complete, which is convenient for the recognition of coupler feature targets. However, the camera can also be installed in other positions where the complete coupler can be photographed.
[0038] In a specific embodiment, the preprocessing includes image enhancement processing. Optionally, the image enhancement processing includes adding Gaussian noise, adjusting brightness and random exposure processing.
[0039] When Gaussian noise is added, the probability density function of the Gaussian noise is:
[0040] (3).
[0041] Where: is the probability density function of Gaussian noise; is a random variable, indicating the value of noise; is the standard deviation, controlling the width of the Gaussian distribution; is the mean;
[0042] It should be noted that the image enhancement processing in the above embodiment is only a preferred image enhancement processing method of the present invention. Other methods in the prior art that can enhance the processing of graphics can also be applied to the present invention. In addition, the purpose of image preprocessing is to improve image quality and enhance useful information for subsequent processing and analysis. In addition to the image enhancement processing in the above embodiment, other preprocessing methods that can achieve this purpose can also be applied to the present invention.
[0043] S2: performing coupler feature target recognition on the preprocessed coupler area image to obtain pixel coordinates of the coupler.
[0044] In a specific embodiment, a deep learning model is used to identify the coupler feature target of the preprocessed coupler area image. It should be noted that target feature recognition of images is a prior art, and the deep learning model of this embodiment is only a preferred recognition method of the present invention. Other methods in the prior art that can identify coupler feature targets from images can also be applied to the present invention, such as traditional feature extraction methods (methods based on manual features), region-based methods (Selective Search, EdgeBoxes, etc.), other methods (Blob analysis method, template matching method, etc.), etc.
[0045] In a specific embodiment, the deep learning model adopts any one of the Yolo series models. Figure 2 As shown, the deep learning model adopts the improved yolov9 network model. Based on the yolov9 network model, the improved yolov9 network model adds the SEAM attention mechanism before its six DETECT output modules and adds the ByteTrack tracking algorithm module after the three DETECT output modules that are always enabled during model training and recognition. Figure 2 In the figure, Conv module is the convolution module, Inputs is the input, RepNCSP is the RepNCSP feature fusion module, Adown is the downsampling module, CBLiner is the connection module between the trunk network and the branch network, CBFuse is the CBFuse feature fusion module, Upsample module is the upsampling module, Concat is the concatenation module, and SPPELAN is the feature pyramid module.
[0046] In the above embodiment, by introducing the SEAM attention mechanism based on the yolov9 network model, the recognition ability of the model can be improved when the target is occluded. The SEAM attention mechanism is a prior art, which uses a deep separable convolution algorithm, which can be divided into two steps: deep convolution and point-by-point convolution. Among them, the deep convolution can be expressed as:
[0047] (4)
[0048] Where: is the input feature map; is the number of channels of the input feature map; For application in The convolution kernel of channels; * is the convolution operation; is the first input feature map channels.
[0049] The point-by-point convolution can be expressed as:
[0050] (5)
[0051] Where: and is the size of the output feature map; is the weight of the convolution kernel.
[0052] In the above embodiment, by introducing the ByteTrack tracking algorithm module based on the yolov9 network model, the ByteTrack tracking algorithm is based on Kalman filtering and Hungarian matching algorithm, which can associate information between different video frames, thereby effectively improving the accuracy of coupler feature target recognition.
[0053] It should be noted that the above-mentioned improved yolov9 network model is only a preferred Yolo series deep learning model of the present invention. Other Yolo series deep learning models that can accurately identify coupler features and other deep learning models (such as CNN, R-CNN, etc.) can also be applied to the present invention.
[0054] S3: Calculate the coupler swing angle according to the pixel coordinates of the coupler.
[0055] In a specific embodiment, the coupler swing angle is calculated by the following formula:
[0056] (1)
[0057] Where: is the hook swing angle; is the number of pixels corresponding to the unit physical size on the horizontal pixel axis; is the number of pixels corresponding to the unit physical size on the pixel vertical axis; is the horizontal coordinate of the pixel coordinate of the coupler; It is half of the total length of the horizontal axis of the image pixels; It is the difference in pixel horizontal coordinates from the center point of the round pin to the center axis of the coupler; The actual longitudinal distance between the camera and the connection point between the coupler and the vehicle body; is the pitch angle of the camera; is the center point of the pin in the world coordinate system Coordinates in the axis direction; is the ordinate of the pixel coordinate of the coupler; It is half of the total length of the vertical axis of the image pixels; is the focal length of the camera.
[0058] In the above embodiment, the calculation formula of the coupler swing angle is derived by the following steps:
[0059] The obtained pixel coordinates of the coupling are transformed. The schematic diagram of the coordinate transformation is as follows: Figure 3 As shown, where O is the camera origin, is the center point of the pin in the camera coordinate system Coordinates in the axis direction; is the center point of the pin in the world coordinate system Direction coordinates; is the center point of the pin in the world coordinate system Direction coordinates; is the center point of the pin in the camera coordinate system Coordinates in the axis direction; is the center point of the pin in the camera coordinate system The coordinates in the axis direction are transformed by the following formula:
[0060] (6)
[0061] (7)
[0062] Where: is the pixel horizontal coordinate of the center point of the circular pin; is the pixel ordinate of the center point of the circle pin.
[0063] After coordinate transformation, the geometric diagram of coupler swing angle calculation is as follows: Figure 4 As shown, P is a point with the same pixel horizontal coordinate as point A and intersecting with the center axis of the coupler, and A is the center point of the round pin; , Respectively represent the pixel horizontal coordinate and pixel vertical coordinate of point P. At this time, the coupler swing angle is calculated by the following formula:
[0064] (8)
[0065] Where: is the center point of the pin in the world coordinate system Direction coordinates.
[0066] Substituting equations (6)-(7) into equation (8) yields the calculation formula for the coupler swing angle shown in equation (1).
[0067] In this embodiment, the swing angle of the coupler can be calculated by obtaining the pixel coordinates of the coupler and the parameters of the camera.
[0068] In another specific embodiment, the coupler swing angle is calculated by the following formula:
[0069] (2)
[0070] Where: is the hook swing angle; , They are the horizontal and vertical coordinates of the pixel coordinates of the coupler respectively; , , All are calibration constants.
[0071] In this embodiment, it is not necessary to obtain the parameters of the camera, and the swing angle of the coupler can be calculated only by obtaining the pixel coordinates of the coupler.
[0072] In a specific embodiment, the calibration constant is obtained by calibration through the following steps: obtaining the measured coupler swing angle and coupler pixel coordinates, using formula (2) to establish equations about the calibration constant under at least three groups of different coupler swing angle conditions; solving the equation to obtain the calibration constant.
[0073] It should be noted that, in the above embodiment, if the camera is at the same position when calibrating the calibration constant, that is, the pixel coordinates of the coupler remain unchanged, then the formula (2) after determining the calibration constant is applicable to the calculation of the coupler swing angle at the same camera position.
[0074] In a specific embodiment, taking a heavy-load train as an example, the heavy-load train coupler swing angle measurement method based on machine vision described in the present invention is used to calculate the identification value of its coupler swing angle, and the true value of the coupler swing angle is obtained by measuring the sensor on the heavy-load train.
[0075] In this embodiment, the camera is installed on the vehicle body just above the coupler, and the original image of the coupler area and the pre-processed image are obtained as follows: Figure 5 The improved yolov9 network model is used to identify the coupler feature target of the preprocessed coupler area image. The result is shown in Figure 6As shown ( Figure 6 The position selected by the square box is the position of the coupler pin, where the left side is the coupler pin of the opposite car body, and the right side is the coupler pin of the current car body), and the pixel coordinates of the coupler are obtained. The coupler swing angle of the heavy-load train is calculated based on the pixel coordinates of the coupler and formula (1).
[0076] In this embodiment, the measurement results of the hook swing angle under the small swing angle condition and the measurement results of the hook swing angle under the large swing angle condition are respectively as follows: Figure 7 and Figure 8 As shown. Figure 7 and Figure 8 It can be seen that the identification value calculated by the present invention is very consistent with the real value measured by the sensor. The present invention can accurately measure the coupler swing angle of a heavy-load train.
[0077] In summary, the present invention can realize high-precision non-contact measurement of the swing angle of the heavy-load train coupler. Compared with the prior art, the present invention has significant progress.
[0078] The above description is only a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.
Claims
1. A method for measuring the swing angle of a heavy-duty train coupler based on machine vision, characterized in that: The following steps are involved: S1: collecting an image of a coupler area of a heavy-load train, and preprocessing the image to obtain a preprocessed coupler area image; S2: performing coupler feature target recognition on the preprocessed coupler region image to obtain pixel coordinates of the coupler; S3: Calculate the coupler swing angle according to the pixel coordinates of the coupler, and the coupler swing angle is calculated by the following formula: （1） Where: is the hook swing angle; is the number of pixels corresponding to the unit physical size on the horizontal pixel axis; is the number of pixels corresponding to the unit physical size on the pixel vertical axis; is the horizontal coordinate of the pixel coordinate of the coupler; It is half of the total length of the horizontal axis of the image pixels; It is the difference in pixel horizontal coordinates from the center point of the round pin to the center axis of the coupler; The actual longitudinal distance between the camera and the connection point between the coupler and the vehicle body; is the pitch angle of the camera; is the center point of the pin in the world coordinate system Coordinates in the axis direction; is the ordinate of the pixel coordinate of the coupler; It is half of the total length of the vertical axis of the image pixels; is the focal length of the camera.
2. The method for measuring the swing angle of a heavy-duty train coupler based on machine vision according to claim 1 is characterized in that: In step S1, the image of the coupler area of the heavy-load train is obtained by a camera installed on the vehicle body just above the coupler.
3. The method for measuring the swing angle of a heavy-duty train coupler based on machine vision according to claim 1, characterized in that: In step S1, the preprocessing includes image enhancement processing.
4. The method for measuring the swing angle of a heavy-duty train coupler based on machine vision according to claim 3 is characterized in that: The image enhancement process includes adding Gaussian noise, adjusting brightness and random exposure processing.
5. The method for measuring the swing angle of a heavy-duty train coupler based on machine vision according to claim 1, characterized in that: In step S2, a deep learning model is used to perform coupler feature target recognition on the preprocessed coupler area image.
6. The method for measuring the swing angle of a heavy-duty train coupler based on machine vision according to claim 5 is characterized in that: The deep learning model adopts any one of the Yolo series models.
7. The method for measuring the swing angle of a heavy-duty train coupler based on machine vision according to claim 6 is characterized in that: The deep learning model adopts an improved yolov9 network model. Based on the yolov9 network model, the improved yolov9 network model adds a SEAM attention mechanism before its six DETECT output modules, and adds a ByteTrack tracking algorithm module after the three DETECT output modules that are always enabled during model training and recognition.
8. The method for measuring the swing angle of a heavy-load train coupler based on machine vision according to any one of claims 1 to 7, characterized in that: In step S3, the coupler swing angle can also be calculated by the following formula: （2） Where: , , All are calibration constants.
9. The method for measuring the swing angle of a heavy-duty train coupler based on machine vision according to claim 8, characterized in that: The calibration constant is obtained by calibration through the following steps: obtaining the measured coupler swing angle and the coupler pixel coordinates, using formula (2) to establish equations about the calibration constant under at least three groups of different coupler swing angle conditions; solving the equation to obtain the calibration constant.