A method and system for detecting abnormality of a railway train wheel set

Abstract

The present invention provides a method and system for detecting abnormality of a track train wheel set, comprising a wheel-rail image acquisition step, for acquiring a wheel-rail image under a straight section; a relative point analysis step, for defining a reference point of a rail, and obtaining a lateral displacement curve and a longitudinal displacement curve according to the distance between a camera and the reference point and the coordinates of the reference point and the rail; a visual acquisition adjustment step, for calculating the acceleration value or deceleration value of the wheelset according to the distance between two adjacent fluctuation units in the longitudinal displacement curve, and controlling the adjustment of a visual camera so that the distance of each fluctuation unit in the longitudinal displacement curve of the reference point is the same; an abnormal screening and matching step, for comparing the steep peak value of each fluctuation unit in the longitudinal displacement curve of the reference point and the vertex value of the lateral displacement curve of the reference point with a preset threshold value, and outputting a wheelset abnormality instruction; the present invention has the advantage that the running state of the wheelset can be comprehensively evaluated by combining the analysis of the acceleration or deceleration value, thereby improving the reliability of the detection result.

Description

Technical Field

[0001] The present invention relates to the technical field of rail vehicle wheel set detection, and more particularly to a rail vehicle wheel set abnormality detection method and system. Background Art

[0002] The wheelset of a rail vehicle is a key component of the train running gear. Its condition directly affects the safety and stability of the train operation. If the wear, cracks, deformation and other defects of the wheelset are not detected and handled in time, serious safety accidents may occur. Therefore, effective inspection of the train wheelset is an important measure to ensure the safety of train operation.

[0003] With the rapid development of high-speed railways, the operating speed of trains is constantly increasing, and the performance requirements for wheelsets are becoming higher and higher. High-speed railway trains need to withstand greater loads and impacts during operation, so the wear and damage of wheelsets will become more serious, resulting in reduced stability of the train during driving, and shaking, tilting, etc.

[0004] In the automated detection of rail vehicle wheelsets, there are currently some technical solutions for wheelset detection, such as the Chinese patent "CN117649437B A rail vehicle wheel-rail lateral displacement measurement method and measurement device", which provides a method for analyzing the lateral offset of point coordinates by using the coordinates of several key points on the vehicle wheel-rail. This method can only analyze the lateral displacement of the wheelset on the wheel-rail, and cannot effectively monitor the damage and vibration of the vehicle wheelset.

[0005] In addition, the Chinese patent disclosed in "CN116534074A A method, device and system for online detection of subway wheel and rail damage" provides a method of collecting the vibration acceleration signal of the subway wheel and rail as a longitudinal acceleration signal through a vibration acceleration sensor, and then processing the longitudinal acceleration signal according to a preset frequency range to obtain the longitudinal displacement of the wheel, and converting the track wear according to the longitudinal displacement. This detection method mainly detects the wear of the wheel and rail, and cannot effectively detect the current driving status of the target train and the wear of the wheelset in a timely manner. Summary of the invention

[0006] In view of the deficiencies in the prior art, an object of the present invention is to provide a method and system for detecting abnormality of a railway train wheelset.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for detecting abnormality of a railway train wheel set comprises the following steps:

[0009] A wheel-rail image acquisition step, obtaining a rail image of a wheel set captured by a visual camera on a straight road section as a wheel-rail image;

[0010] In the relative point analysis step, one or more reference points of the rail are respectively mapped in the wheel-rail image, and a continuous point position curve diagram is obtained according to the distance calculation between the camera and the reference point and the coordinate calculation between the reference point and the rail, wherein the point position curve diagram includes a lateral displacement curve of the reference point and a longitudinal displacement curve of the reference point;

[0011] A visual acquisition adjustment step, in which an acceleration value or a deceleration value of the wheelset is calculated according to the distance between two adjacent fluctuation units in the reference point longitudinal displacement curve, and a visual camera is controlled to adjust based on the acceleration value or the deceleration value so that the distance of each fluctuation unit in the reference point longitudinal displacement curve is the same;

[0012] The abnormal screening and matching step compares the steep peak value of each fluctuation unit in the reference point longitudinal displacement curve and the vertex value of the reference point lateral displacement curve with the preset threshold values, and outputs the wheelset abnormal instruction and the corresponding wheelset driving status according to the comparison results.

[0013] Furthermore, the visual acquisition adjustment step is connected to a camera shooting frequency table in an external adjustment system library, wherein the camera shooting frequency table includes a camera shooting frequency and a train speed, and the camera shooting frequency corresponds to the train speed one by one. The current real-time wheelset speed is calculated according to the acceleration value or the deceleration value, and the train speed corresponding to the real-time wheelset speed is indexed in the camera shooting frequency table, and the camera shooting frequency corresponding to the speed is extracted as the frequency collected by the current visual camera.

[0014] Furthermore, the visual acquisition adjustment step controls the visual camera to move in the opposite direction or in the same direction according to the acceleration value or deceleration value of the wheelset to offset the acceleration value or deceleration value of the wheelset.

[0015] Furthermore, it also includes a wheel-rail anomaly elimination step, defining the wheelset of the first car of the train as the initial judgment wheelset, calibrating the point curve diagram of the initial judgment wheelset when an abnormal result occurs, and the corresponding rail section as the calibration section, respectively obtaining the point curve diagram of the wheelset of each subsequent car passing through the calibration section, and again comparing the steep peak value of each fluctuation unit in each point curve diagram and the vertex value of the reference point lateral displacement curve with the preset threshold value, and judging the number of abnormalities in the corresponding wheelset according to the comparison result, if the number of abnormalities is greater than the threshold, outputting the rail abnormality instruction, otherwise, outputting the wheelset abnormality instruction.

[0016] Furthermore, the relative item point analysis step obtains a wheel-rail image of any rail, draws a horizontal straight line transversely to the reference point on the rail as a reference baseline, and maps the reference baseline to the wheel-rail image;

[0017] The abnormal screening and matching step calculates the distance between each reference baseline and the edge of the wheelset image in the continuous wheel-rail images as the reference distance. If the reference distance changes, a vehicle body tilt instruction is output, otherwise, a normal instruction is output.

[0018] Furthermore, in the abnormal screening and matching step, a vehicle body inclination curve is drawn according to the reference distance in two adjacent wheel-rail images. If the vehicle body inclination curve has one fluctuation unit or multiple irregular fluctuation units, a rail abnormality instruction is output; if the vehicle body inclination curve has a regular cyclic fluctuation unit, a wheelset abnormality instruction is output.

[0019] Furthermore, it also includes a wheelset abnormality verification step. When a wheelset abnormality instruction is obtained, a wheel-rail image of another rail is obtained, the reference distance calculation and the body inclination curve are repeated, and the body inclination curves of the two rails are compared. If the fluctuation unit is located in the same wheel-rail image, an accurate abnormality screening instruction is output, otherwise, an fluctuation unit repeated analysis instruction is output.

[0020] Furthermore, it also includes a vehicle body tilt feedback step, calculating the steep peak value and the trough value of the fluctuation unit in the vehicle body tilt curve, calculating the steep peak value and the trough value through trigonometric function or externally adjusting the system library index to obtain the vehicle body tilt angle, comparing the vehicle body tilt angle with a threshold, and outputting monitoring instructions and danger instructions according to the comparison result.

[0021] Furthermore, in the visual acquisition adjustment step, the shooting angle of the camera is calibrated according to the position of the reference point mapped in the wheel-rail image.

[0022] A rail train wheelset abnormality detection system, comprising:

[0023] The wheel-rail image acquisition module acquires the rail image of the wheelset captured by the visual camera on the straight section as the wheel-rail image;

[0024] The relative point analysis module maps one or more reference points of the rail into the wheel-rail image, calculates the distance between the camera and the reference point, and calculates the coordinates of the reference point and the wheel-rail to obtain a continuous point curve diagram, wherein the point curve diagram includes a lateral displacement curve of the reference point and a longitudinal displacement curve of the reference point;

[0025] A visual acquisition adjustment module, which calculates the acceleration value or deceleration value of the wheelset according to the distance between two adjacent fluctuation units in the reference point longitudinal displacement curve, and controls the adjustment of the visual camera based on the acceleration value or deceleration value, so that the distance of each fluctuation unit in the reference point longitudinal displacement curve is the same;

[0026] The abnormal screening and matching module compares the steep peak value of each fluctuation unit in the reference point longitudinal displacement curve and the vertex value of the reference point lateral displacement curve with the preset threshold value, and outputs the wheelset abnormal instruction and the corresponding wheelset driving status according to the comparison result.

[0027] The beneficial effects of the present invention are as follows: by analyzing the relative position of the rail and the camera through visually captured wheel-rail images, a point curve diagram that is easy to analyze is formulated; and then according to the adjustment of the visual shooting, the distance of each fluctuation unit in the point curve diagram is made the same when the train is in an accelerating or decelerating state, which facilitates the analysis of the fluctuation unit to output whether the wheelset is abnormal. The present invention not only simply focuses on the longitudinal displacement and lateral displacement of the wheelset, but also combines the analysis of the acceleration / deceleration values, so as to comprehensively evaluate the running status of the wheelset. At the same time, the dynamic characteristics of the train during driving are further understood, and the reliability of the detection results is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is the overall flow chart of the present invention;

[0029] Figure 2 It is the point curve diagram in the present invention;

[0030] Figure 3 is a vehicle body tilt determination flow chart in the present invention;

[0031] Figure 4 is a distribution diagram of the vehicle bottom vision cameras in the present invention;

[0032] Figure 5 It is the vehicle body inclination curve diagram in the present invention. DETAILED DESCRIPTION

[0033] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The same parts are represented by the same reference numerals. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the accompanying drawings, and the words "bottom surface" and "top surface", "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0034] Since the current wheelset detection is only to analyze the lateral displacement and longitudinal displacement of the wheelset on the wheel-rail when the wheelset moves to judge the wear of the rails, and does not effectively monitor the damage and vibration of the vehicle wheelset, the present invention designs such a track train wheelset abnormality detection method, such as Figure 1 As shown, the following steps are included:

[0035] Wheel-rail image acquisition steps, such as Figure 4 As shown, a visual camera and an auxiliary light source are provided on the wheelset side of the train. Due to the interference of wheelset movement and train inertia when traveling on a curve, in order to eliminate the above interference, the visual camera continuously captures the images of the rails and the wheelset above them as wheel-rail images when the train is traveling on a straight section. In the initial state, the installation position of the camera is adjusted, and the auxiliary light source is appropriate. The wheelset image is clearer, and the relative position of the rails and the wheelset can be clearly distinguished.

[0036] In the relative point analysis step, a key reference point on the rail (such as a gauge point or a center point on the side of the rail, etc.) is selected, where the reference point can be one or more. The present invention takes a reference point as an example, and then uses image processing technology, such as feature extraction and matching algorithm, to map and accurately locate the reference point in the track image. The actual distance between the camera and the reference point can be determined according to the pixel distance calculation method of the visual image. The change in the distance can reflect the lateral offset distance of the wheelset on the track. A lateral displacement curve is obtained by drawing the distance at a certain time, wherein the abscissa is time and the ordinate is the distance value. The offset distance is then obtained according to the relative position (coordinate calculation) of the reference point and the wheel-rail (rail) in the wheel-rail image. The change in the offset distance can reflect the longitudinal offset distance of the wheelset on the track. A longitudinal displacement curve is obtained by drawing the offset distance at a certain time, wherein the abscissa is time and the ordinate is the offset distance value. The above-mentioned lateral displacement curve and longitudinal displacement curve are simultaneously drawn in a point position curve diagram, and are two curves corresponding to each other on the time axis, as shown in FIG. Figure 2 As shown, the gentle curve on the top is the lateral displacement curve, and the undulating unit on the bottom is the longitudinal displacement curve.

[0037] Visual acquisition adjustment step: Since the train may have an accelerated driving state, a uniform speed driving state, and a decelerated driving state, when the camera shooting frequency is a fixed frequency under the uniform speed driving state, the reference point in the captured image will present a fluctuating curve due to the shaking of the train itself, but under the accelerated driving state and the decelerated driving state, the time interval of the fluctuating curve will become narrower or longer, that is, under the accelerated driving state, the fluctuating curve segment is relatively dense, and under the decelerated driving state, the fluctuating curve segment is relatively elongated, and the analysis accuracy will be reduced when the fluctuating curve segment is dense or elongated, and the accuracy of abnormal detection of wheelsets under the accelerated and decelerated states is higher than that under the uniform speed state. Therefore, the acceleration value or deceleration value of the wheelset is calculated by the distance between two adjacent fluctuating units in the longitudinal displacement curve, and the visual camera adjustment is controlled based on the acceleration value or deceleration value, so that the distance of each fluctuating unit in the longitudinal displacement curve of the reference point is the same, which is convenient for the subsequent abnormal analysis of the wheelset under the accelerated driving state and the decelerated driving state.

[0038] In the abnormal screening and matching step, the adjusted reference point longitudinal displacement curve is further analyzed to identify the steep peak in the fluctuation unit. The steep peak reflects that the wheelset is subjected to abnormal impact or vibration. At the same time, the vertex value of the reference point lateral displacement curve is checked. These values ​​reflect the offset of the wheelset in the lateral direction. Excessive offset may be caused by wheelset wear. The steep peak value is compared with the threshold value N1, and the vertex value is compared with the threshold value N2. If any one exceeds the threshold, the wheelset abnormal instruction is output, and the section where the abnormality is located is extracted, and the corresponding driving status is attached to facilitate the driver to make timely judgments.

[0039] Through the above analysis method, real-time monitoring and abnormality detection of the train wheelset operating status can be achieved, thereby improving the safety and reliability of train operation.

[0040] Embodiment 1:

[0041] like Figure 1 As shown, the visual acquisition adjustment step is to connect the camera shooting frequency table in the external adjustment system library, which includes a camera shooting frequency column and a train speed column, wherein each train speed corresponds to a camera shooting frequency, and the specific corresponding means include experimental means or a large amount of historical data stacked according to the deep learning model to build, and calculate the wheel set speed (real-time train speed) in real time, calculate the current real-time speed by the speed calculation formula through the current acceleration value or deceleration value, and index the train speed corresponding to the real-time wheel set speed in the camera shooting frequency table, and extract the camera shooting frequency corresponding to the speed as the frequency of the current visual camera acquisition. This kind of visual adjustment method can quickly control the camera through the frequency-speed relationship without changing the camera installation structure, so as to achieve the shooting purpose and avoid delay.

[0042] Embodiment 2:

[0043] like Figure 1 As shown, the visual acquisition adjustment steps are as follows: a slide rail is provided at the bottom of the carriage along the length direction of the carriage, and a visual camera is slidably connected to the slide rail. The sliding of the visual camera on the slide rail is controlled according to the acceleration value or deceleration value of the wheelset to offset the acceleration value or deceleration value of the wheelset. If the wheelset is in an accelerated state, the visual camera moves on the slide rail in an accelerated direction opposite to the direction of travel of the train. If the wheelset is in a decelerated state, the visual camera moves on the slide rail in an accelerated direction in the same direction as the direction of travel of the train. At this time, the relative movement of the camera and the slide rail always maintains a uniform speed. Due to the limited length of the slide rail, only wheel-rail images of a partial length of acceleration or deceleration can be obtained. The length of the fluctuation unit of the acceleration part and the deceleration part in the drawn point curve diagram is shorter, but the abnormal conditions of the wheelset under the acceleration and deceleration states can also be analyzed.

[0044] Embodiment 3:

[0045] On the basis of Example 1 or Example 2, since abnormal wheelset or rail abnormality will cause abnormal changes in the wheel-rail image, a wheel-rail abnormality elimination step is also included. The first step is to select an initial judgment wheelset: define the wheelset of the first carriage in the driving direction as the initial judgment wheelset, which is the first wheelset to contact the rail, and its data can be used as a reference for subsequent analysis;

[0046] The second step is to calibrate the abnormal point curve: when the point curve of the initial judgment wheelset shows abnormal results (such as exceeding the normal range, abnormal fluctuations, etc.), the rail section that the wheelset passes through is calibrated as the "calibration section". This calibration section is the key area for subsequent analysis;

[0047] The third step is to obtain the point curve diagram of the subsequent carriages: obtain the point curve diagram of the wheelset of each subsequent carriage passing through the calibrated section of rail;

[0048] Step 4: Fluctuation unit analysis: In each point curve diagram, identify and extract each fluctuation unit, extract the steep peak value of each fluctuation unit (i.e., the point with the largest fluctuation amplitude) and the vertex value of the reference point lateral displacement curve (if applicable), and compare them with the preset threshold value N3 respectively. If it is greater than the threshold value N3, the rail abnormality instruction is output, indicating that the rails in the calibration section may have defects or problems. Otherwise, the wheelset abnormality instruction is output, which may be a problem with the wheelset itself, such as wear, imbalance, etc.

[0049] The fifth step is the feedback step. When it is an abnormal rail instruction, the information of the calibrated section of the rail is transmitted to the railway road administration system for marking. When it is an abnormal wheelset instruction, the abnormal wheelset information (wheelset number of the carriage) is transmitted to the railway maintenance system for marking.

[0050] When the wheelset or rail is abnormal, the vehicle body will tilt. When the vehicle is running at high speed, the tilt of the vehicle body will seriously affect the safety of the vehicle. Figure 3 As shown in the figure, relative point analysis steps are used to define a reference baseline: two rails are needed to carry a set of wheels. Select any rail to obtain its wheel-rail image. The defined reference point is assumed to be the center point of the side of the rail. A horizontal straight line is drawn along the length direction of the rail as the reference baseline. The reference baseline is mapped to the first wheel-rail image and used as a reference for measuring the inclination of the vehicle body.

[0051] Abnormal screening and matching steps, the first step is to calculate the reference distance: each wheel-rail image includes a section of rail, so continuous wheel-rail images are spliced ​​to obtain continuous rails. However, when analyzing each wheel-rail image, the distance between each reference baseline and the edge of the wheelset image is called the reference distance. Under normal circumstances, the vehicle body is not tilted, and the fluctuation of this reference line is a fluctuation curve with the vibration of the wheelset. However, when the vehicle body is tilted, the shooting angle of the camera changes, and the fluctuation curve of the reference baseline will also change, that is, the wheel-rail image will shift downward or upward;

[0052] The second step is to draw the body tilt curve: Draw the body tilt curve according to the reference distance in two adjacent wheel-track images, such as Figure 5 As shown;

[0053] The third step is to analyze the body tilt curve: If the body tilt curve has one fluctuation unit or multiple irregular fluctuation units (such as Figure 5 A-1 and B-1 in the figure), then the rail abnormality instruction is output. If there is a regular cyclic fluctuation unit in the body tilt curve (such as Figure 5 If the output is C-1, the wheelset abnormal instruction is output.

[0054] like Figure 3As shown in the figure, theoretically, when the vehicle body tilts, one side should be high and the other side should be low. Then, the abnormal wheel set verification step is also included. When the abnormal wheel set instruction is obtained, the wheel-rail image corresponding to the currently detected wheelset and located on the other side of the train (i.e., the other rail) needs to be obtained first. This is to verify the authenticity of the wheelset abnormality from another angle, eliminate the misjudgment caused by the problem of a single rail, and repeat the calculation of the reference distance for the wheel-rail image on the other rail. This includes defining the reference baseline, measuring the distance between the edge of the wheelset image and the baseline, and recording the reference distance value in each frame of the image. The corresponding vehicle body tilt curve is drawn using the reference distance data obtained from the other rail. This curve will reflect the change of the vehicle body tilt when the train is running on the other rail. The vehicle body tilt curves of the two rails are compared. If the fluctuation unit is located in the same wheel-rail image, it means that similar vehicle body tilt changes are detected on two different rails at the same time or at a similar time. This increases the authenticity and accuracy of wheelset abnormalities. The system should output "accurate abnormality screening instructions" to confirm that there is an abnormality in the wheelset. If the fluctuation unit is inconsistent in the body inclination curves of the two rails, or it appears repeatedly in multiple inspections of the same rail but there is no corresponding change in the other side of the rail, this may indicate a misjudgment caused by a rail problem, sensor error or data processing error. At this time, the system should output "fluctuation unit repeated analysis instructions" to prompt further inspection and analysis; it can more comprehensively evaluate the status of wheelsets and rails, reduce the possibility of misjudgment and missed judgment, and improve the safety and reliability of train operation.

[0055] The system also includes a vehicle body tilt feedback step, which calculates the steep peak value and the trough value of the fluctuation unit in the vehicle body tilt curve, obtains the vehicle body tilt angle by calculating the steep peak value and the trough value through trigonometric functions or by externally adjusting the system library index, compares the vehicle body tilt angle with a threshold value, and outputs monitoring instructions and danger instructions according to the comparison result.

[0056] Since the camera's shooting angle will shift when the train is running at high speed for a long time, the visual acquisition adjustment step calibrates the camera's shooting angle according to the position of the reference point mapped in the wheel-rail image. Currently, the correction of the camera's shooting angle is an existing technology, which can be mainly corrected by the coordinate point of the reference point in the wheel-rail image.

[0057] In view of the above detection method, a wheelset abnormality detection system is provided, including

[0058] The wheel-rail image acquisition module acquires the rail image of the wheelset captured by the visual camera on the straight section as the wheel-rail image;

[0059] The relative point analysis module maps one or more reference points of the rail into the wheel-rail image, calculates the distance between the camera and the reference point, and calculates the coordinates of the reference point and the wheel-rail to obtain a continuous point curve diagram, wherein the point curve diagram includes a lateral displacement curve of the reference point and a longitudinal displacement curve of the reference point;

[0060] A visual acquisition adjustment module, which calculates the acceleration value or deceleration value of the wheelset according to the distance between two adjacent fluctuation units in the reference point longitudinal displacement curve, and controls the adjustment of the visual camera based on the acceleration value or deceleration value, so that the distance of each fluctuation unit in the reference point longitudinal displacement curve is the same;

[0061] The abnormal screening and matching module compares the steep peak value of each fluctuation unit in the reference point longitudinal displacement curve and the vertex value of the reference point lateral displacement curve with the preset threshold value, and outputs the wheelset abnormal instruction and the corresponding wheelset driving status according to the comparison result.

[0062] Correspondingly, it also includes a wheel-rail anomaly elimination module, a wheelset anomaly verification module, a vehicle body tilt feedback module, a command interaction module and a data transmission module. The functions of the above modules correspond to the steps in the method. Its purpose is to automatically acquire images - automatically analyze and process images - automatically judge - automatically make decisions, combined with analyzing acceleration / deceleration values, so as to comprehensively evaluate the operating status of the wheelset. At the same time, it further understands the dynamic characteristics of the train during travel and improves the reliability of the detection results.

[0063] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for detecting abnormality of a railway train wheel set, characterized in that: The steps include: A wheel-rail image acquisition step, obtaining a rail image of a wheel set captured by a visual camera on a straight road section as a wheel-rail image; In the relative point analysis step, one or more reference points of the rail are respectively mapped in the wheel-rail image, and a continuous point position curve diagram is obtained according to the distance calculation between the camera and the reference point and the coordinate calculation between the reference point and the rail, wherein the point position curve diagram includes a lateral displacement curve of the reference point and a longitudinal displacement curve of the reference point; A visual acquisition adjustment step, in which the acceleration value or deceleration value of the wheelset is calculated according to the distance between two adjacent fluctuation units in the reference point longitudinal displacement curve, and the shooting frequency or accelerated movement of the visual camera is controlled based on the acceleration value or deceleration value, so that the distance of each fluctuation unit in the reference point longitudinal displacement curve is the same; The abnormal screening and matching step compares the steep peak value of each fluctuation unit in the reference point longitudinal displacement curve and the vertex value of the reference point lateral displacement curve with the preset threshold values, and outputs the wheelset abnormal instruction and the corresponding wheelset driving status according to the comparison results.

2. According to claim 1, a method for detecting abnormality of a railway train wheel set is characterized in that: The visual acquisition adjustment step is connected to a camera shooting frequency table in an external adjustment system library, wherein the camera shooting frequency table includes a camera shooting frequency and a train speed, wherein the camera shooting frequency corresponds to the train speed in a one-to-one manner, and the current real-time wheelset speed is calculated according to the acceleration value or the deceleration value, and the train speed corresponding to the real-time wheelset speed is indexed in the camera shooting frequency table, and the camera shooting frequency corresponding to the speed is extracted as the frequency collected by the current visual camera.

3. According to claim 1, a method for detecting abnormality of a railway train wheel set is characterized in that: The visual acquisition adjustment step controls the visual camera to move in the opposite direction or in the same direction according to the acceleration value or deceleration value of the wheelset to offset the acceleration value or deceleration value of the wheelset.

4. A method for detecting abnormality of a railway train wheel set according to claim 2 or 3, characterized in that: The method also includes a wheel-rail anomaly elimination step, defining the wheelset of the first carriage of the train as the wheelset for initial judgment, calibrating the point position curve diagram of the wheelset for initial judgment when an abnormal result occurs as the calibration section, obtaining the point position curve diagram of the wheelset of each subsequent carriage passing through the calibration section, and again comparing the steep peak value of each fluctuation unit in each point position curve diagram and the vertex value of the lateral displacement curve of the reference point with a preset threshold value, judging the number of anomalies in the corresponding wheelset according to the comparison result, and outputting a rail anomaly instruction if the number of anomalies is greater than the threshold, otherwise, outputting a wheelset anomaly instruction.

5. A method for detecting abnormality of a railway train wheel set according to claim 2 or 3, characterized in that: The relative point analysis step obtains a wheel-rail image of any rail, draws a horizontal straight line transversely to a reference point on the rail as a reference baseline, and maps the reference baseline to the wheel-rail image; The abnormal screening and matching step calculates the distance between each reference baseline and the edge of the wheelset image in the continuous wheel-rail images as the reference distance. If the reference distance changes, a vehicle body tilt instruction is output, otherwise, a normal instruction is output.

6. A method for detecting abnormality of a railway train wheel set according to claim 5, characterized in that: The abnormal screening and matching step draws a vehicle body inclination curve according to the reference distance in two adjacent wheel-rail images. If the vehicle body inclination curve has one fluctuation unit or multiple irregular fluctuation units, a rail abnormality instruction is output; if the vehicle body inclination curve has a regular cyclic fluctuation unit, a wheelset abnormality instruction is output.

7. A method for detecting abnormality of a railway train wheel set according to claim 6, characterized in that: It also includes a wheelset abnormality verification step. When a wheelset abnormality instruction is obtained, a wheel-rail image of another rail is obtained, the reference distance calculation and the body inclination curve are repeated, and the body inclination curves of the two rails are compared. If the fluctuation unit is located in the same wheel-rail image, an accurate abnormality screening instruction is output, otherwise, an fluctuation unit repeated analysis instruction is output.

8. A method for detecting abnormality of a railway train wheelset according to claim 7, characterized in that: The system also includes a vehicle body tilt feedback step, in which the steep peak value and the trough value of the fluctuation unit are calculated in the vehicle body tilt curve, the steep peak value and the trough value are calculated by trigonometric function or an external system library index to obtain the vehicle body tilt angle, the vehicle body tilt angle is compared with a threshold value, and monitoring instructions and danger instructions are output according to the comparison result.

9. A method for detecting abnormality of a railway train wheelset according to claim 1, characterized in that: The visual acquisition adjustment step calibrates the shooting angle of the camera according to the position of the reference point mapped in the wheel-rail image.

10. A railway train wheel set abnormality detection system, characterized in that: include The wheel-rail image acquisition module acquires the rail image of the wheelset captured by the visual camera on the straight section as the wheel-rail image; The relative point analysis module maps one or more reference points of the rail into the wheel-rail image, calculates the distance between the camera and the reference point, and calculates the coordinates of the reference point and the wheel-rail to obtain a continuous point curve diagram, wherein the point curve diagram includes a lateral displacement curve of the reference point and a longitudinal displacement curve of the reference point; A visual acquisition adjustment module, which calculates the acceleration value or deceleration value of the wheelset according to the distance between two adjacent fluctuation units in the reference point longitudinal displacement curve, and controls the shooting frequency or accelerated movement of the visual camera based on the acceleration value or deceleration value, so that the distance of each fluctuation unit in the reference point longitudinal displacement curve is the same; The abnormal screening and matching module compares the steep peak value of each fluctuation unit in the reference point longitudinal displacement curve and the vertex value of the reference point lateral displacement curve with the preset threshold value, and outputs the wheelset abnormal instruction and the corresponding wheelset driving status according to the comparison result.

Images