Railway track safety monitoring method based on image recognition

Abstract

The invention discloses a railway track safety monitoring method based on image recognition, and belongs to the technical field of railway track monitoring, and the method comprises the steps: S1, obtaining the sectional inertia moment of a to-be-monitored steel rail; s2, pasting a plurality of reflective identification sheets; s3, arranging a high-frequency camera; s4, preliminarily segmenting a plurality of reflective mark areas; s5, pixel points of the connection areas of the adjacent reflective identification areas are accurately segmented, and the relative positions of the reflective identification sheets are stored and recorded according to the coordinates of the centroid pixel points of the reflective identification sheets; s6, generating a displacement data graph of the steel rails between the adjacent sleepers; the actual maximum displacement, the maximum train stress, the wheel-rail force, the vibration acceleration and the wheel load shedding rate of the steel rails between the adjacent sleepers are obtained; and S8, monitoring the monitored wheel-rail force, vibration acceleration and wheel load shedding rate in real time. The monitoring method disclosed by the invention is simple to operate, low in cost, high in safety and high in measurement precision, and the comprehensiveness and the reliability of an evaluation result are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of railway track monitoring, and in particular to a railway track safety monitoring method based on image recognition. Background Art

[0002] With the rapid development of railways, the requirements for the stability of railway operation are becoming higher and higher. Affected by the vibration of trains passing, mechanical wear, weather, rain and snow, etc., deformations that cannot be completely eliminated are likely to occur during train operation, and in severe cases may even cause track breakage. Therefore, in order to ensure the safe operation of railways, it is necessary to monitor the dynamic displacement safety of railway tracks.

[0003] At present, the main methods for monitoring railway track displacement are: track inspection vehicles, which are equipped with various sensors and measuring equipment, and can measure various parameters of the track, but this method is expensive and can only be performed during the skylight period; laser measurement technology, which uses laser rangefinders and other equipment to remotely scan and measure the track to obtain various parameters of the track. This method has the advantages of non-contact and high precision, but its cost is high, and the error of dynamic monitoring when the train is running is large; installing sensors on the track, such as vibration sensors, which monitor the vibration of the train when passing through the track, thereby inferring the dynamic displacement of the track. Its advantages are low cost and simple operation, but the sensors are prone to damage or tilt during the monitoring process, resulting in deviations in the measurement results, and corresponding data cables need to be installed on the line for data transmission, which has high labor costs. Summary of the invention

[0004] The purpose of the present invention is to provide a railway track safety monitoring method based on image recognition to solve the problems of high cost and large error in existing railway track dynamic monitoring.

[0005] The technical solution of the present invention to solve the above technical problems is as follows:

[0006] A railway track safety monitoring method based on image recognition comprises the following steps:

[0007] S1: perform profile measurement on the monitoring section of the rail to be monitored to obtain the cross-sectional inertia moment of the rail to be monitored;

[0008] S2: Paste a number of reflective identification sheets on the monitoring section of the rail to be monitored;

[0009] S3: Place a high-frequency camera so that it faces the rails. When the train runs in the monitoring section, the high-frequency camera takes pictures of the rails.

[0010] S4: Preliminarily segmenting a number of corresponding reflective logo areas according to the reflective logo sheet;

[0011] S5: accurately segmenting the pixel points of the connecting area of ​​the adjacent reflective logo areas, and storing and recording the relative position of the reflective logo piece according to the coordinates of the centroid pixel point of the reflective logo piece;

[0012] S6: generating a displacement data graph of rails between adjacent sleepers according to the displacement change data of the centroid pixel point during the train operation;

[0013] S7: According to the displacement data graph and the cross-sectional inertia moment of the rail, the actual maximum displacement of the rails between adjacent sleepers, the maximum train stress, the wheel-rail force, the vibration acceleration and the wheel load reduction rate are obtained when the train is running in the monitoring section;

[0014] S8: Real-time monitoring of the wheel-rail force, vibration acceleration and wheel load reduction rate.

[0015] Furthermore, in the above step S1, the section for profile measurement is the rail section between adjacent sleepers, and the average value of the moments of inertia of all the sections measured is taken.

[0016] Furthermore, in the above step S2, the reflective identification sheets are pasted at the middle and bottom of the rail section between adjacent sleepers, and at the fasteners on the top of the sleepers, and all the reflective identification sheets are located on the same side of the rail.

[0017] Furthermore, in the above step S4, the Laplace Gaussian operator algorithm is used for preliminary segmentation, and the specific formula is as follows:

[0018]

[0019] Among them, ΔG(x i ,y i ) represents the Laplace operator processing of the i-th pixel point in the adjacent reflective identification area through the two-dimensional Gaussian distribution formula, x i Indicates the horizontal coordinate of the ith pixel in the adjacent reflective identification area relative to the base point, y i It represents the vertical coordinate of the ith pixel point in the adjacent reflective identification piece area relative to the base point, and σ represents the standard deviation of the distance between the pixels in the adjacent reflective identification piece area.

[0020] Furthermore, the formula for accurately segmenting the pixels of the connecting area of ​​adjacent reflective logo areas is:

[0021]

[0022] Wherein, D represents the connection area of ​​adjacent reflective logo areas, H represents the distance between the pixel points of the connection area of ​​adjacent reflective logo areas and one of the reflective logo areas, S represents the distance between the pixel points of the connection area of ​​adjacent reflective logo areas and the other reflective logo area, and d jrepresents the distance between the jth pixel point of the connecting area of ​​adjacent reflective logo areas and the pixel points of two adjacent reflective logo areas, Δd represents the adjustment distance for dividing the area after judging whether a certain pixel point of the connecting area of ​​adjacent reflective logo areas belongs to a certain reflective logo area, K represents the pixel definition of the high-frequency camera, ΔK represents the pixel definition automatically adjusted by the high-frequency camera according to the demand, and x j represents the horizontal coordinate of the jth pixel point on the connecting area dividing line of adjacent reflective logo areas, Δx represents the automatically adjusted value of the horizontal coordinate of the point on the connecting area dividing line of adjacent reflective logo areas, and y j represents the vertical coordinate of the jth pixel point on the connecting area dividing line of adjacent reflective logo areas, Δy represents the automatic adjustment value of the vertical coordinate of the point on the connecting area dividing line of adjacent reflective logo areas, σ represents the standard deviation of the distance between the connecting area pixel point of the adjacent reflective logo area and the pixel points of two adjacent reflective logo areas, and Δσ represents the automatic adjustment value of the standard deviation of the distance between the connecting area pixel point of the adjacent reflective logo area and the pixel points of two adjacent reflective logo areas.

[0023] Furthermore, the centroid of the reflective identification sheet is:

[0024]

[0025] Where z is 1 or 2, target represents the white pixel area, background represents the black pixel background area, and n represents the number of all pixels in the adjacent reflective logo area;

[0026] When z is 1, x i1 Indicates the horizontal coordinate of the pixel point in one of the two adjacent reflective logo areas, y i1 represents the vertical coordinate of a pixel point in one of the two adjacent reflective identification areas, f(x i1 ,y i1 ) indicates that the binary image is in (x i1 ,y i1 ), x 01 Indicates the horizontal coordinate of the centroid of one of the reflective identification patch areas, y 01 The vertical coordinate representing the centroid of one of the retroreflective identification patch areas;

[0027] When z is 2, x i2 Indicates the horizontal coordinate of the pixel point in the other reflective identification area between two adjacent reflective identification areas, y i2 represents the vertical coordinate of the pixel point in the other reflective identification area of ​​the adjacent two reflective identification areas, f(x i2 ,y i2) indicates that the binary image is in (x i2 ,y i2 ), x 02 Indicates the horizontal coordinate of the centroid of the other retro-reflective patch area, y 02 Indicates the vertical coordinate of the centroid of another retroreflective identification patch area.

[0028] Furthermore, in the above step S7, the actual maximum displacement d 实际 for:

[0029]

[0030] Wherein, d is the maximum vertical displacement of the rail section between adjacent sleepers, obtained from the displacement data graph, x1 is the maximum vertical displacement of the reflective identification piece on the top of one of the adjacent sleepers, and x2 is the maximum vertical displacement of the reflective identification piece on the top of the other of the adjacent sleepers;

[0031] The maximum train stress P is:

[0032]

[0033] Where, E is the elastic modulus of the rail, and I is the section moment of inertia of the rail;

[0034] The vibration acceleration A is:

[0035]

[0036] Among them, F is the vibration frequency, AMP is the vibration amplitude, and g is the gravitational acceleration;

[0037] Wheel load reduction rate:

[0038]

[0039] Among them, P1 is the heavier wheel weight of the left and right wheels of the train, P2 is the lighter wheel weight of the left and right wheels of the train, and ΔP is the wheel weight reduction of the wheel on the load-reducing side. It is the average static wheel weight of the wheels on the unloaded and loaded sides.

[0040] The present invention has the following beneficial effects:

[0041] (1) The present invention is easy to operate. It only requires attaching the reflective identification sheet to the designed position of the rail and then taking a quick photo with a high-frequency camera. The reflective identification sheet is relatively easy to replace. If the identification sheet is tilted or damaged during train operation, it can be quickly replaced during the window time.

[0042] (2) The present invention has low cost. The main monitoring equipment is only a reflective identification sheet and a high-frequency camera. The production cost of the reflective identification sheet is low, and the high-frequency camera can be repeatedly used in multiple different scenes.

[0043] (3) The present invention has high security. The reflective identification sheet in the field does not need a data line for data transmission, eliminating the possibility of safety hazards in the data line. Secondly, the high-frequency camera can be manually adjusted and placed in a safe area outside the field, which greatly ensures the safety of monitoring personnel and equipment.

[0044] (4) The present invention introduces more relevant parameters such as wheel load reduction rate to judge the safety of rails through image data processing, thereby improving the comprehensiveness and reliability of the judgment results.

[0045] (5) Based on the segmentation using the Laplace operator algorithm, the present invention further judges the pixel points of two adjacent areas and introduces an automatically adjusted logic algorithm to perform more accurate graphic segmentation, thereby greatly improving the accuracy of the graphic segmentation technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the flow of the railway track safety monitoring method based on image recognition of the present invention;

[0047] Figure 2 It is a schematic diagram of pasting the reflective identification sheet of the present invention;

[0048] Figure 3 This is a schematic diagram of the high-frequency camera of the present invention when taking pictures.

[0049] In the figure: 1-sleeper; 2-fastener; 3-rail; 4-reflective identification sheet; 5-high-frequency camera. DETAILED DESCRIPTION

[0050] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0051] Please refer to Figure 1 This embodiment provides a railway track safety monitoring method based on image recognition, comprising the following steps:

[0052] S1: Perform profile measurement on the monitoring section of the monitored rail to obtain the cross-sectional inertia moment of the monitored rail. The section for profile measurement is the rail section between adjacent sleepers. When there are multiple sleepers in the monitoring section, multiple groups of cross-sectional inertia moments will be obtained. At this time, the cross-sectional inertia moments of all groups are averaged to obtain the cross-sectional inertia moment of the rail. In order to reduce the error, the maximum and minimum values ​​of the cross-sectional inertia moment can also be removed and then the average value is taken to obtain the cross-sectional inertia moment of the rail.

[0053] S2: Paste several reflective identification sheets on the monitoring section of the rail 3 to be monitored. Figure 2 As shown, the reflective identification sheets 4 are pasted at the middle and bottom of the section of the rail 3 between the adjacent sleepers 1, and at the fastener 2 at the top of the sleeper 1, and all the reflective identification sheets 4 are located on the same side of the rail 3. The displacement of the reflective identification sheet 4 at the fastener 2 is the displacement of the corresponding sleeper 1.

[0054] In order to reduce the possibility of the reflective identification sheet 4 falling or tilting due to unstable installation, the installation base surface should be polished with fine sandpaper before installation until the surface of the patch is flat, and then the surface of the rail 3 should be gently wiped with a cotton swab dipped in 95% anhydrous ethanol to remove powder and oil on the surface of the rail 3. When pasting the reflective identification sheet, ensure that there are no bubbles.

[0055] S3: Place the high-frequency camera so that it faces the rails. When the train runs in the monitoring section, the high-frequency camera takes pictures of the monitoring section of the rails to be monitored. Figure 3 As shown, the high-frequency camera 5 should be placed in an area with a wide field of view and no obstructions outside the field. Under the premise of ensuring that the high-frequency camera 5 can completely capture the monitoring section of the steel rail 3 to be monitored, the high-frequency camera 5 should be aligned with the steel rail 3 as much as possible. The installed high-frequency camera 5 is connected to the data storage device, the data converter, and the PC computer in sequence. The data storage device is used to store data (such as photos, centroid coordinates, etc.), and the data converter is used to convert data (such as converting the relative position data of the centroid into relative displacement data, etc.) and transmit the converted data to the PC computer.

[0056] S4: Preliminarily segmenting a number of corresponding reflective logo areas according to the reflective logo sheet.

[0057] In this embodiment, the Laplace Gaussian operator algorithm is used for preliminary segmentation, and the specific formula is as follows:

[0058]

[0059] Among them, ΔG(x i ,y i ) represents the Laplace operator processing of the i-th pixel point in the adjacent reflective identification area through the two-dimensional Gaussian distribution formula, x i Indicates the horizontal coordinate of the ith pixel in the adjacent reflective identification area relative to the base point, y i It represents the vertical coordinate of the ith pixel point in the adjacent reflective identification piece area relative to the base point, and σ represents the standard deviation of the distance between the pixels in the adjacent reflective identification piece area.

[0060] S5: accurately segmenting the pixel points of the connecting area of ​​the adjacent reflective logo areas, and storing and recording the relative position of the reflective logo piece according to the coordinates of the centroid pixel point of the reflective logo piece.

[0061] In this embodiment, the formula for accurately segmenting the pixel points of the connecting area of ​​adjacent reflective logo areas is:

[0062]

[0063] Wherein, D represents the connection area of ​​adjacent reflective logo areas, H represents the distance between the pixel point of the connection area of ​​adjacent reflective logo areas and the centroid of one of the reflective logo areas, S represents the distance between the pixel point of the connection area of ​​adjacent reflective logo areas and the centroid of the other reflective logo area, and d j represents the distance between the jth pixel point of the connecting area of ​​adjacent reflective logo areas and the pixel points of two adjacent reflective logo areas, Δd represents the adjustment distance for dividing the area after judging whether a certain pixel point of the connecting area of ​​adjacent reflective logo areas belongs to a certain reflective logo area, K represents the pixel definition of the high-frequency camera, ΔK represents the pixel definition automatically adjusted by the high-frequency camera according to the demand, and x j represents the horizontal coordinate of the jth pixel point on the connecting area dividing line of adjacent reflective logo areas, Δx represents the automatically adjusted value of the horizontal coordinate of the point on the connecting area dividing line of adjacent reflective logo areas, and y j represents the vertical coordinate of the jth pixel point on the connecting area dividing line of adjacent reflective logo areas, Δy represents the automatic adjustment value of the vertical coordinate of the point on the connecting area dividing line of adjacent reflective logo areas, σ represents the standard deviation of the distance between the connecting area pixel point of the adjacent reflective logo area and the pixel points of two adjacent reflective logo areas, and Δσ represents the automatic adjustment value of the standard deviation of the distance between the connecting area pixel point of the adjacent reflective logo area and the pixel points of two adjacent reflective logo areas.

[0064] After the image is segmented by the Laplace operator algorithm, further judgment is made on the pixel points of the connection area of ​​the two adjacent connection areas, and the distance, centroid distance, and standard deviation of the distance from the pixel point on the connection area dividing line of the two adjacent connection areas to the pixel points of the two adjacent connection areas are compared. If the distance, centroid distance, and standard deviation of the distance from the pixel point on the connection area dividing line to the pixel point of one area of ​​the two adjacent areas are greater than the distance, centroid distance, and standard deviation of the distance from the pixel point of the other area, the pixel point is divided into a smaller distance identification piece area. If any of the conditions does not meet the requirements, the pixel point is re-divided and segmented, the position of the segmentation boundary line is automatically adjusted, and the relevant parameters are automatically adjusted and iterated. The step size of each iteration is about one percent of the initial parameter. After that, the distance, centroid distance, and standard deviation of the distance from the pixel point on the connection area dividing line to the two adjacent areas are recalculated again, and compared again until the constraint conditions are met, otherwise other parameters are replaced and the operation is repeated. The centroid position of the area after accurate segmentation is accurate, thereby ensuring the accuracy of the displacement data obtained.

[0065] In this embodiment, the centroid of the reflective identification sheet is:

[0066]

[0067] Where z is 1 or 2, target represents the white pixel area, background represents the black pixel background area, and n represents the number of all pixels in the adjacent reflective logo area;

[0068] When z is 1, x i1 Indicates the horizontal coordinate of the pixel point in one of the two adjacent reflective logo areas, y i1 represents the vertical coordinate of a pixel point in one of the two adjacent reflective identification areas, f(x i1 ,y i1 ) indicates that the binary image is in (x i1 ,y i1 ), x 01 Indicates the horizontal coordinate of the centroid of one of the reflective identification patch areas, y 01 The vertical coordinate representing the centroid of one of the retroreflective identification patch areas;

[0069] When z is 2, x i2 Indicates the horizontal coordinate of the pixel point in the other reflective identification area between two adjacent reflective identification areas, y i2 represents the vertical coordinate of the pixel point in the other reflective identification area of ​​the adjacent two reflective identification areas, f(x i2 ,yi2 ) indicates that the binary image is in (x i2 ,y i2 ), x 02 Indicates the horizontal coordinate of the centroid of the other retro-reflective patch area, y 02 Indicates the vertical coordinate of the centroid of another retroreflective identification patch area.

[0070] S6: generating a displacement data graph of rails between adjacent sleepers according to the displacement change data of the centroid pixel point during the train operation;

[0071] S7: According to the displacement data diagram and the cross-sectional inertia moment of the rail, the actual maximum displacement of the rails between adjacent sleepers, the maximum train stress, the wheel-rail force, the vibration acceleration and the wheel load reduction rate are obtained when the train is running in the monitoring section.

[0072] Actual maximum displacement d 实际 for:

[0073]

[0074] Wherein, d is the maximum vertical displacement of the rail section between adjacent sleepers, obtained from the displacement data graph, x1 is the maximum vertical displacement of the reflective identification piece on the top of one of the adjacent sleepers, and x2 is the maximum vertical displacement of the reflective identification piece on the top of the other of the adjacent sleepers;

[0075] The maximum train stress P is:

[0076]

[0077] Where, E is the elastic modulus of the rail, and I is the section moment of inertia of the rail;

[0078] The vibration acceleration A is:

[0079]

[0080] Among them, F is the vibration frequency, AMP is the vibration amplitude, and g is the gravitational acceleration;

[0081] Wheel load reduction rate:

[0082]

[0083] Among them, P1 is the heavier wheel weight of the left and right wheels of the train, P2 is the lighter wheel weight of the left and right wheels of the train, and ΔP is the wheel weight reduction of the wheel on the load-reducing side. It is the average static wheel weight of the wheels on the unloaded and loaded sides.

[0084] S8: Real-time monitoring of wheel-rail force, vibration acceleration and wheel load reduction rate, and alarms when abnormal data occurs. At the same time, the photos taken by the high-frequency camera when abnormal data occurs can be extracted and stored to analyze the causes of abnormal data. The real-time monitoring data can also be used as a training set to analyze and predict the track displacement and related data in the monitoring area.

[0085] The railway track safety monitoring method based on image recognition in this embodiment is simple to operate, low in cost and highly safe. It introduces more relevant parameters such as wheel weight reduction rate to judge the safety of the rails, thereby improving the comprehensiveness and reliability of the judgment results. At the same time, the segmentation accuracy of the reflective mark area is high, thereby improving the measurement accuracy.

[0086] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A railway track safety monitoring method based on image recognition, characterized in that: The following steps are involved: S1: perform profile measurement on the monitoring section of the rail to be monitored to obtain the cross-sectional inertia moment of the rail to be monitored; S2: Paste a number of reflective identification sheets on the monitoring section of the rail to be monitored; S3: Place a high-frequency camera so that it faces the rails. When the train runs in the monitoring section, the high-frequency camera takes pictures of the rails. S4: Preliminarily segmenting a number of corresponding reflective logo areas according to the reflective logo sheet; S5: accurately segmenting the pixel points of the connecting area of ​​the adjacent reflective logo areas, and storing and recording the relative position of the reflective logo piece according to the coordinates of the centroid pixel point of the reflective logo piece; S6: generating a displacement data graph of rails between adjacent sleepers according to the displacement change data of the centroid pixel point during the train operation; S7: According to the displacement data graph and the cross-sectional inertia moment of the rail, the actual maximum displacement of the rails between adjacent sleepers, the maximum train stress, the wheel-rail force, the vibration acceleration and the wheel load reduction rate are obtained when the train is running in the monitoring section; S8: Real-time monitoring of the wheel-rail force, vibration acceleration and wheel load reduction rate.

2. The railway track safety monitoring method based on image recognition according to claim 1 is characterized in that: In step S1, the section for profile measurement is the rail section between adjacent sleepers, and the average value of all the measured moments of inertia of the sections is taken.

3. The railway track safety monitoring method based on image recognition according to claim 1 is characterized in that: In step S2, the reflective identification sheets are pasted at the middle and bottom of the rail section between adjacent sleepers, and at the fasteners on the top of the sleepers, and all the reflective identification sheets are located on the same side of the rail.

4. The railway track safety monitoring method based on image recognition according to claim 1 is characterized in that: In step S4, the Laplace Gaussian operator algorithm is used for preliminary segmentation, and the specific formula is as follows: Among them, ΔG(x i ,y i ) represents the Laplace operator processing of the i-th pixel point in the adjacent reflective identification area through the two-dimensional Gaussian distribution formula, x i Indicates the horizontal coordinate of the ith pixel in the adjacent reflective identification area relative to the base point, y i It represents the vertical coordinate of the ith pixel point in the adjacent reflective identification piece area relative to the base point, and σ represents the standard deviation of the distance between the pixels in the adjacent reflective identification piece area.

5. The railway track safety monitoring method based on image recognition according to claim 4 is characterized in that: The formula for accurately segmenting the pixels of the connected area of ​​adjacent reflective logo areas is: Wherein, D represents the connection area of ​​adjacent reflective logo areas, H represents the distance between the pixel point of the connection area of ​​adjacent reflective logo areas and the centroid of one of the reflective logo areas, S represents the distance between the pixel point of the connection area of ​​adjacent reflective logo areas and the centroid of the other reflective logo area, and d j represents the distance between the jth pixel point of the connecting area of ​​adjacent reflective logo areas and the pixel points of two adjacent reflective logo areas, Δd represents the adjustment distance for dividing the area after judging whether a certain pixel point of the connecting area of ​​adjacent reflective logo areas belongs to a certain reflective logo area, K represents the pixel definition of the high-frequency camera, ΔK represents the pixel definition automatically adjusted by the high-frequency camera according to the demand, and x j represents the horizontal coordinate of the jth pixel point on the connecting area dividing line of adjacent reflective logo areas, Δx represents the automatically adjusted value of the horizontal coordinate of the point on the connecting area dividing line of adjacent reflective logo areas, and y j represents the vertical coordinate of the jth pixel point on the connecting area dividing line of adjacent reflective logo areas, Δy represents the automatic adjustment value of the vertical coordinate of the point on the connecting area dividing line of adjacent reflective logo areas, σ represents the standard deviation of the distance between the connecting area pixel point of the adjacent reflective logo area and the pixel points of two adjacent reflective logo areas, and Δσ represents the automatic adjustment value of the standard deviation of the distance between the connecting area pixel point of the adjacent reflective logo area and the pixel points of two adjacent reflective logo areas.

6. The railway track safety monitoring method based on image recognition according to claim 5 is characterized in that: Centroid of reflective logo piece: Where z is 1 or 2, target represents the white pixel area, background represents the black pixel background area, and n represents the number of all pixels in the adjacent reflective logo area; When z is 1, x i1 Indicates the horizontal coordinate of the pixel point in one of the two adjacent reflective logo areas, y i1 represents the vertical coordinate of a pixel point in one of the two adjacent reflective identification areas, f(x i1 ,y i1 ) indicates that the binary image is in (x i1 ,y i1 ), x 01 Indicates the horizontal coordinate of the centroid of one of the reflective identification patch areas, y 01 The vertical coordinate representing the centroid of one of the retroreflective identification patch areas; When z is 2, x i2 Indicates the horizontal coordinate of the pixel point in the other reflective identification area between two adjacent reflective identification areas, y i2 represents the vertical coordinate of the pixel point in the other reflective identification area of ​​the adjacent two reflective identification areas, f(x i2 ,y i2 ) indicates that the binary image is in (x i2 ,y i2 ), x 02 Indicates the horizontal coordinate of the centroid of the other retro-reflective patch area, y 02 Indicates the vertical coordinate of the centroid of another retro-reflective identification patch area.

7. The railway track safety monitoring method based on image recognition according to claim 1 is characterized in that: In step S7, the actual maximum displacement d 实际 for: Wherein, d is the maximum vertical displacement of the rail section between adjacent sleepers, obtained from the displacement data graph, x1 is the maximum vertical displacement of the reflective identification piece on the top of one of the adjacent sleepers, and x2 is the maximum vertical displacement of the reflective identification piece on the top of the other of the adjacent sleepers; The maximum train stress P is: Where, E is the elastic modulus of the rail, and I is the section moment of inertia of the rail; The vibration acceleration A is: Among them, F is the vibration frequency, AMP is the vibration amplitude, and g is the gravitational acceleration; Wheel load reduction rate: Among them, P1 is the heavier wheel weight of the left and right wheels of the train, P2 is the lighter wheel weight of the left and right wheels of the train, and ΔP is the wheel weight reduction of the wheel on the load-reducing side. It is the average static wheel weight of the wheels on the unloaded and loaded sides.

Images