A railway slope monitoring method and system based on sensor fusion and a target body

Abstract

The present application relates to a kind of railway slope monitoring method, system and target based on sensor fusion.The method comprises: layout fusion monitoring target and obtain its spatial coordinates and visual identification;Establish radar image element coordinates, video holder parameter and the multi-sensor mapping relationship of target;Through radar real-time monitoring target displacement and reflection intensity, trigger video check when threshold value is exceeded;Camera periodically visits target, and triggers radar precision measurement when identifying change;Decision alarm based on two-way verification result.The system contains radar, video sensor, edge computing terminal and early warning platform.The target is composed of support, horizontal / tilt angle adjustable support, radar corner reflector and chessboard visual board.The present application reduces environmental interference false alarm, improves monitoring accuracy and reliability through radar and video data fusion and two-way triggering mechanism, and is suitable for complex railway slope scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of railway slope monitoring, in particular to a railway slope monitoring method and system based on sensor fusion and a target body. BACKGROUND

[0002] Currently, railway slope monitoring usually adopts GNSS Beidou high-precision positioning system. Although it has high-precision positioning capability, it still faces many challenges in complex railway slope environment. For example, bad weather can affect the signal, resulting in discontinuous or fluctuating monitoring data; it is difficult to install and maintain in steep slope areas, and it cannot fully capture small deformation and has monitoring blind area. As an active monitoring means, slope monitoring radar has non-contact monitoring capability. This monitoring method can monitor the slope in real time without interfering with the normal operation of the railway. At the same time, radar monitoring is not limited by light and weather conditions, and can maintain stable monitoring effect even at night or in bad weather. In order to improve the accuracy of slope radar monitoring of the slope body, corner reflectors are usually arranged at the positions of interest of the slope to enhance the reflection intensity of the radar wave, so as to accurately monitor the displacement. However, in the railway application scenario, the high-frequency vibration caused by high-speed trains (freight or passenger) and the complex environment of railway line engineering, simply relying on radar single means monitoring cannot intuitively and completely accurately reflect the change information of the slope. When the slope environment changes, such as personnel construction, slope inspection, external interference, etc., it will cause the corner of the slope and the monitoring target of the concerned part to be blocked, causing false alarm of radar and false reporting. And the structure and installation flexibility and practicality of the traditional corner reflector in the railway application scenario and the long-term stability of use are difficult to meet the actual demand. SUMMARY

[0003] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, one object of the present application is to propose a railway slope monitoring method and system based on sensor fusion and a target body, which captures the small deformation information of the slope target body through the radar system, simultaneously uses the video device to record the slope monitoring target device and the appearance change of the slope in real time, and through the fusion analysis of the data of the two, can more effectively identify the potential disaster risk, reduce the false reporting rate, and improve the monitoring efficiency and accuracy. And the present application relies on the fusion monitoring method to design a target body. This fusion technology method and new target body not only overcome the limitations of traditional single monitoring means, but also provide a new solution for the safety management of railway slope.

[0004] In order to solve the above problems, the present application provides a railway slope monitoring method based on sensor fusion, comprising the following steps:

[0005] Step 1: planning and laying the fusion monitoring target body based on the geological survey results;

[0006] Step 2: Obtain the precise spatial coordinates and visual identification information of the fusion monitoring target, and generate QP values ​​based on the chessboard visual board on the target to form a mapping relationship table (ID, Xn, Yn, Zn, QP).

[0007] Step 3: Sensor installation and debugging. Based on the on-site field of view, determine the optimal installation points for the radar and video surveillance; adjust the azimuth and elevation angles of the radar so that its beam can effectively cover all monitored targets; adjust the orientation and focal length of the camera so that it can clearly capture the target image.

[0008] Step 4: Establish multi-sensor collaborative mapping relationships;

[0009] Step 4.1: Calibrate Radar Image Pixel Coordinates: Read the precise coordinates of all targets and the radar's own pose parameters in the radar system, and perform coordinate transformation to establish a mapping table of "target ID - Radar Image Pixel Coordinates (ID, Radar_I, Radar_J)".

[0010] Step 4.2: Establish a video sensor mapping table: Control the pan-tilt camera to point it at each target and the monitoring target of interest in turn; for each target, manually or automatically adjust the focus until the chessboard visual board image is clearest, and then record the azimuth angle of the pan-tilt at this time. Pitch angle and focal length parameters This is bound to the target and the target of interest ID, forming a mapping table of "target and target of interest ID - gimbal preset position parameters" (ID, , , Once completed, the camera captures a clear image of each target, which is then stored in the database as the initial background image Pic0.

[0011] Step 4.3: Establish a radar-visual-target mapping table: Establish the mapping relationship between radar video targets using the unique ID of the target or object, forming a unique mapping relationship between each sensor based on the ID (ID). n Xn, Yn, Zn, QP n Radar_I n Radar_J n , n , n , n , , , (); where ID is the target code, X, Y, Z are the spatial coordinates of the target, QP is the target visual chessboard code, and Radar_I and Radar_J are the radar image metadata coordinates. For the target distance, The target azimuth angle, For the target scattering intensity, , , The preset azimuth, pitch, and focal length parameters of the camera relative to the target, where n represents the nth target;

[0012] Step 5: The radar actively monitors and analyzes the displacement and reflection intensity changes of the target in real time; in conjunction with the periodic inspection and trigger-based verification by the camera, the changes in the appearance of the target are identified through image comparison; when either the radar or the camera detects an anomaly in the target first, the other detection method is immediately triggered for collaborative verification, and finally, a graded warning or prompt is issued based on the dual verification results.

[0013] Preferably, Step 4.1 includes the following sub-steps:

[0014] Step 4.1.1: Convert the world coordinate system to the radar's local rectangular coordinate system;

[0015] First, to describe the positional offset of the target relative to the radar phase center, the translation vector T needs to be obtained:

[0016] ;

[0017] in, Geodetic coordinates of the target measured by RTK / total station; : Geodetic coordinates of the radar antenna phase center;

[0018] Rotation matrix The parameter matrix is ​​used to rotate the world coordinate system (x, y, z) to the radar installation azimuth deflection direction; where a R is the radar's azimuth offset angle:

[0019] ;

[0020] Coordinate transformation result: The coordinates of the target point in the radar's local rectangular coordinate system are obtained. ;

[0021] ;

[0022] Step 4.1.2: Convert the radar's local rectangular coordinate system to the radar's polar coordinate system;

[0023] Through geometric calculations, the slant range and azimuth angle used by the radar to sense the target are obtained;

[0024] In the formula The straight-line spatial distance between the geodetic coordinates of the radar antenna phase center and the geodetic coordinates of the target center;

[0025] Azimuth ;

[0026] Step 4.1.3: Convert the radar polar coordinate system to radar cell index;

[0027] Distance Gate Index :

[0028] ;

[0029] in, The minimum detection range set for the radar; Radar range resolution, ,in At the speed of light, For signal bandwidth;

[0030] Azimuth Unit Index :

[0031] ;

[0032] in, The minimum azimuth angle for radar detection; Angular resolution of the radar;

[0033] Calculate the range-angle unit that each target should theoretically appear in the radar image to establish a mapping table of "target ID - radar image unit coordinates (ID, Radar_I, Radar_J)";

[0034] This is used to calibrate radar data from relative measurements to absolute geographic coordinates, so that the displacement data monitored by the radar directly corresponds to the spatial displacement in the real world; the mapping relationship between radar image pixel coordinates (Radar_I, Radar_J) and target distance is also considered. Azimuth and scattering The value is unique.

[0035] Preferably, in Step 5, the edge end processes the radar echo signal in real time, performing two analyses for each target pixel region in the mapping table:

[0036] A. Displacement Analysis: Calculate the phase change of the current pixel and inversely determine the cumulative displacement of the target relative to its initial position. R;

[0037] B. Reflection Intensity Analysis: Monitor the change in radar echo intensity of this pixel. ;

[0038] Then the displacement is compared with the preset multi-level early warning thresholds. The RJ sensor is compared, including a noticeable value of 5mm, a warning value of 10mm, and an alarm value of 20mm, while simultaneously determining the reflection intensity. Does a mutation exist? If any condition exceeds the limit, the system immediately marks the target as "abnormal".

[0039] When the radar detects an anomaly in the target, it triggers a verification process. At the same time, it sends a command to the camera through the mapping table, requiring it to immediately capture video image information of the abnormal target. Then, based on the video image detection AI algorithm, the target is visually identified, triggering a visual verification of the camera image.

[0040] Preferably, in Step 5, when the camera first detects an anomaly in the target, the radar verifies the anomaly. The radar receives the verification request from the video system and performs a rapid scan and data analysis on the designated target. If the radar verification result shows that the displacement and reflection intensity of the target are within limits, the system determines that this is a false alarm from the video and does not issue an alarm, but only records a "video detection change prompt" in the system log for management personnel to review. If the radar verification confirms that the displacement or intensity of the target has exceeded the limits, it indicates that the dual verification by radar and video is successful, and the risk of actual deformation of the slope is extremely high. The system immediately generates alarm information including alarm level, target location, displacement data, and on-site pictures, and issues it through various means such as sound and light, SMS, and platform push to remind relevant personnel to take emergency measures.

[0041] A system used in a sensor fusion-based railway slope monitoring method includes a data interconnection terminal layer, a network transmission layer, and an application service layer, wherein:

[0042] The terminal layer includes slope radar, cameras, and edge computing terminals. The slope radar is used to collect slope surface displacement data; the cameras are used to collect slope image information; and the edge computing terminals are used to analyze, verify, and identify radar data and video data.

[0043] Network transport layer: Used to transmit data collected by the terminal layer;

[0044] Application service layer: This includes a radar video surveillance and early warning platform, which is used to receive, process, and display monitoring data and generate early warning information.

[0045] A target used in a sensor fusion-based railway slope monitoring method includes a support, an angle adjustment frame, a radar corner reflector, and a checkerboard vision board. The top of the support rotates to support the angle adjustment frame on a horizontal plane. The pitch angle adjustment frame supports the radar corner reflector and the checkerboard vision board. The checkerboard vision board has multiple squares arranged in alternating black and white colors.

[0046] Preferably, the bottom of the angle adjustment frame is fixed with a horizontal angle indexing flange, the middle of the angle adjustment frame is horizontally inserted with a rotating shaft, and the top of the angle adjustment frame is made into a pitch angle indexing plate with a top arc surface; the edge of the pitch angle indexing plate is provided with multiple horizontally penetrating indexing holes.

[0047] The horizontal angle-split flange is adjusted horizontally on the top plane of the support and bolted together.

[0048] One end of the radar corner reflector and the chessboard vision plate are rotatably mounted on the rotating shaft, while the other end of the radar corner reflector is pinned to the indexing hole, and the other end of the chessboard vision plate is attached to the radar corner reflector for pitch angle positioning constraint.

[0049] The advantages of this invention compared to the prior art are:

[0050] This invention, in the field of railway slope monitoring, utilizes a microwave slope monitoring radar, a camera, and a novel target as the terminal layer. Through a sensor-level deep fusion logic identification method based on radar video, a non-contact, high-precision monitoring system for railway slopes based on a radar-video target is formed. Compared to existing monitoring methods, this system achieves planar, non-contact, and highly interference-resistant high-precision monitoring of railway slopes.

[0051] In the field of railway slope monitoring, this invention employs microwave slope monitoring radar and camera sensors as the terminal foundation to design and establish a two-way sensor-level fusion detection and discrimination method for detecting minute changes in the monitored target. Through cross-verification of radar (precise quantification) and video (intuitive identification), false alarms caused by environmental interference from a single sensor are effectively filtered out. Two-way triggered active detection greatly improves the reliability of alarm information and reduces false alarms caused by interference. The entire process forms a complete intelligent closed loop of "perception-trigger-verification-decision," reducing manual intervention and improving the accuracy of early warning compared to existing monitoring methods.

[0052] This invention, in the field of railway slope monitoring, combines the aforementioned sensor fusion methods with railway application scenarios to design a novel target structure. The horizontal angle adjustment frame, radar corner reflector, and checkerboard vision plate are adjusted by rotating the horizontal angle indexing flange. The pitch angle of the radar corner reflector and checkerboard vision plate is adjusted by inserting positioning pins into the indexing holes at different positions, ensuring that the target is not prone to loosening or movement during long-term monitoring. The modular design improves efficiency in production and transportation. This target design can simultaneously adapt to the monitoring needs of radar and video sensors, and provides reliable stability even in strong winds and heavy rain. The overall structure can adapt to complex field environments for extended periods. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart of the method of the present invention;

[0055] Figure 2 This is a system architecture diagram of the present invention;

[0056] Figure 3 This is a schematic diagram of the three-dimensional structure of the target in this invention;

[0057] Figure 4 This is a side view of the target in this invention;

[0058] In the diagram: 1-Radar corner reflector; 2-Checkerboard vision panel; 3-Index hole; 4-Angle adjustment bracket; 5-Horizontal angle indexing flange; 6-Support; 7-Rotating shaft; 8-Pitch angle indexing plate. Detailed Implementation

[0059] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0060] The present invention provides a railway slope monitoring method based on sensor fusion, comprising the following steps:

[0061] Phase 1: System Initialization and Calibration (Step 1 ~ Step 4)

[0062] Step 1: Railway geological survey and target deployment planning

[0063] During implementation, geological engineers first conduct a detailed survey of the target railway slope. By analyzing the rock and soil structure, the orientation of weak surfaces, and hydrological conditions, they identify the potential sliding modes and directions of the slope. Based on this conclusion, the locations of monitoring targets are planned at key displacement characterization points of the slope (such as the rear edge tensile cracking zone, the middle shear zone, and the front edge bulging zone), forming a monitoring network that can comprehensively reflect the stability state of the slope.

[0064] Step 2: Target Installation and Spatial Information Acquisition

[0065] The corner reflector device of this invention is transported to the planned design location for installation. During installation, it is necessary to ensure that its reinforced uprights are firmly buried in the ground or anchored to stable bedrock. After installation, a high-precision measuring instrument (such as a small RTK or total station) is used to obtain millimeter-level accuracy, measuring and recording the three-dimensional spatial coordinates (Xn, Yn, Zn) of the center point of each target reflector plate. Based on the black and white markings of the target's checkerboard pattern, each visual checkerboard grid consists of black and white colors. If the number of checkerboard grids is NumG, then the final result can be... Each target has a unique visual identification code, and the visual identification code of each target ID can be recorded using a QP value, where QP contains a table relationship between the grid number and its color value. In this way, during target installation and measurement, the target ID, target spatial coordinates, and target code can be mapped to form a mapping table (ID, Xn, Yn, Zn, QP).

[0066] Step 3: Sensor Installation and Debugging

[0067] Based on the on-site field of view, determine the optimal installation points for the radar and video surveillance. Adjust the radar's azimuth and elevation angles to ensure its beam effectively covers all monitored targets. Adjust the camera's orientation and focal length to clearly capture images of the targets. Similarly, using a total station (or a small RTK) or similar equipment, accurately measure the coordinates, azimuth, and elevation angles of the radar antenna phase center and the camera's optical center, and record these measurements into the system.

[0068] Step 4: Establishing Multi-Sensor Collaborative Mapping Relationships

[0069] Step 4.1: Radar Image Element Coordinate Calibration: Read the precise coordinates of all targets and the radar's own pose parameters from the radar system. This is achieved through the coordinate transformation model in Step 4.1.3.

[0070] Step 4.1.1: World Coordinate System → Radar Local Cartesian Coordinate System

[0071] Translation vector T: describes the positional offset of the target relative to the radar phase center.

[0072]

[0073] in, Geodetic coordinates of the target (measured using RTK / total station). : Geodetic coordinates of the phase center of the radar antenna.

[0074] Rotation matrix Align the axes of the world coordinate system with the radar's viewpoint. (The remaining text appears to be incomplete and requires further context.) a R is the azimuth offset of the radar installation.

[0075] ;

[0076] Coordinate transformation result: The coordinates of the target point in the radar's local rectangular coordinate system are obtained. .

[0077]

[0078] Step 4.1.2: Radar Local Cartesian Coordinate System → Radar Polar Coordinate System

[0079] Through geometric calculations, two core physical quantities used by radar to sense targets are obtained: slant range and azimuth.

[0080] In the formula The slant range is the straight-line distance between the geodetic coordinates of the radar antenna phase center and the geodetic coordinates of the target center.

[0081] Azimuth

[0082] Step 4.1.3: Radar Polar Coordinate System → Radar Pixel Index

[0083] Distance Gate Index :

[0084]

[0085] in, The minimum detection range set for the radar. Radar range resolution ( , At the speed of light, (This refers to the signal bandwidth).

[0086] Azimuth Unit Index :

[0087]

[0088] in, : The minimum azimuth angle for radar detection. : Radar angular resolution.

[0089] Calculate the range-angle unit (i.e., pixel) that each target should theoretically appear in the radar image. Establish a mapping table of "target ID - radar image pixel coordinates (ID, Radar_I, Radar_J)".

[0090] This process calibrates radar data from relative measurements to absolute geographic coordinates, enabling radar-monitored displacement data to directly correspond to real-world spatial displacement. Specifically, it establishes a unique mapping relationship between the target distance and the radar image metadata coordinates (Radar_I, Radar_J). Azimuth and scattering value.

[0091] Step 4.2: Establishing the video sensor mapping table:

[0092] By controlling the pan-tilt camera, it is aligned sequentially with each target and the area of ​​interest for monitoring. For each target, the focus is manually or automatically adjusted until the image (especially the chessboard visual board 2) is at its clearest, and then the azimuth angle of the pan-tilt is set. Pitch angle Focal length parameters Save this information and bind it to the target and target of interest IDs to form a mapping table of "target and target of interest IDs - gimbal preset position parameters" (ID, , , Once completed, the camera captures a clear image of each target, which is then stored in the database as the initial background image Pic0.

[0093] Step 4.3: Establishment of the Radar-Visual-Target Mapping Table: A unique mapping relationship (ID, Xn, Yn, Zn, QP) exists when selecting a target or object; a mapping relationship (ID, Radar_I, Radar_J) exists during radar detection; and a mapping relationship (ID, Xn, Yn, Zn, QP) exists during video camera detection. , , Mapping relationships. A mapping relationship can be established for radar video targets using their unique IDs, forming a unique mapping relationship between each sensor based on the ID, i.e., (ID...). n Xn, Yn, Zn, QP n Radar_I n Radar_J n , n , n , n , , , Where ID is the target code, X, Y, Z are the spatial coordinates of the target, QP is the target visual chessboard code, and Radar_I and Radar_J are the radar image primitive coordinates. For the target distance, The target azimuth angle, For the target scattering intensity, , , These are the preset azimuth, pitch, and focal length parameters of the camera relative to the target. In these parameters, the subscript n represents the nth target.

[0094] Phase Two: Routine Monitoring and Collaborative Judgment (Step 5)

[0095] Step 5: Active Radar Monitoring

[0096] After the system starts, the radar scans the monitoring area at a set frequency (e.g., once every 5 minutes). The edge processing unit processes the radar echo signal in real time, performing two analyses for each target pixel region in the target ID-radar image pixel coordinate (ID, Radar_I, Radar_J) mapping table:

[0097] A. Displacement Analysis: Calculate the phase change of the current pixel to deduce the cumulative displacement of the target relative to its initial position. R.

[0098] B. Reflection Intensity Analysis: Monitor the change in radar echo intensity of this pixel. .

[0099] The displacement is compared with the preset multi-level early warning thresholds. The reflection intensity is compared with that of RJ sensors (e.g., 5mm for attention, 10mm for warning, 20mm for alarm, etc.) and simultaneously determined. The system checks for sudden changes (such as a significant weakening or strengthening of the echo due to the target tilting or being obstructed; this can be determined by setting a threshold, such as a change of 20 dB or more). If any condition exceeds the limit, the system immediately marks the target as "abnormal" and triggers an alarm process. Simultaneously, it sends a command to the video system via a mapping table, requesting it to immediately view the video image information of the abnormal target. The system then uses video image detection (AI) algorithms to visually identify the target, triggering a visual verification step.

[0100] Periodic inspection and trigger-based verification of cameras

[0101] Without human intervention, the camera automatically calls the preset position parameters in the mapping table to perform a periodic inspection of all targets according to a preset time plan (e.g., once every hour). During each inspection, it automatically captures the current image PicI and compares it with Pic0 in the database.

[0102] Unidentified Changes: If the image recognition algorithm (which can use the traditional frame difference method or an AI-based image change detection model) determines that the image has not changed significantly, the system will not alarm and can update Pic0 with the current image PicI to eliminate the influence of slow environmental changes such as seasons and lighting.

[0103] Change Detection: If the algorithm detects changes such as tilting, displacement, occlusion, or cracks appearing in the background, it immediately triggers the "active radar detection reporting" process. The video system sends the target's ID to the radar system, requesting the radar to immediately perform precise measurement and analysis of that specific target.

[0104] Radar review and final decision

[0105] After receiving the video verification request, the radar will quickly scan and analyze the designated target within 2 seconds.

[0106] If the radar verification results show that the displacement and reflection intensity of the target are within the limits, the system will determine that this is a false alarm (possibly caused by changes in light and shadow or temporary obstruction), and will not issue an alarm, but will only record a "video detection change prompt" in the system log for administrators to review.

[0107] If radar verification confirms that the target's displacement or strength has exceeded the limit, it indicates successful dual verification by radar and video, and the risk of actual slope deformation is extremely high. The system immediately generates alarm information (including alarm level, target location, displacement data, on-site images, etc.) and releases it through various means such as sound and light, SMS, and platform push notifications to remind relevant personnel to take emergency measures.

[0108] The system used in the railway slope monitoring method based on sensor fusion of the present invention consists of a terminal layer, a network transmission layer, and application services.

[0109] 1. Terminal layer:

[0110] (1) Slope radar: A ground-based railway geological slope monitoring radar that conforms to the "Radio Radio Management Regulations" band shall be selected, which has millimeter-level displacement monitoring accuracy and minute-level acquisition frequency. The radar shall be installed on a stable base facing the monitored slope. Its installation position (latitude, longitude, elevation), orientation (azimuth), and elevation angle shall be accurately measured and recorded by equipment such as a small RTK or total station.

[0111] (2) Video sensor / camera (can be integrated or separate): Select a high-definition network intelligent PTZ camera with automatic zoom function, which supports preset position settings and PTZ control. Its installation position and attitude parameters also need to be accurately measured and recorded.

[0112] (3) Integrated monitoring targets: The integrated high-stability spatial adjustment monitoring and calibration device (i.e., corner reflector device) of this invention is deployed as cooperative targets at key points on the slope (such as potential slip surface exits and displacement-sensitive areas). The deployment location of the target needs to be scientifically planned based on the geological survey conclusions. And when the installation is completed, its information is recorded and a mapping relationship is established with the checkerboard visual board.

[0113] 2. Network Transmission Layer: At the railway site, raw data collected by radar and video equipment is transmitted in real time to the data processing and decision-making center server deployed in the work area or cloud via industrial switches, fiber optics, or 5G / 4G wireless networks.

[0114] 3. Application Service Layer: Dedicated data fusion and processing software is deployed to execute the initialization calibration and routine monitoring logic described below. A web or mobile visualization interface is provided to users to display real-time monitoring data, alarm information, historical trend curves, and on-site video footage.

[0115] The target used in the railway slope monitoring method based on sensor fusion of the present invention is implemented through the following steps for installation and adjustment: Taking the installation of the corner reflector device of the present invention on a railway slope as an example:

[0116] 1. Site Selection and Pole Installation: Excavate a foundation pit at the planned location, pour concrete to form a foundation, and install and fix support 6 to the foundation to ensure its verticality and stability. The design of the support greatly enhances the pole's resistance to overturning and impact after backfilling.

[0117] 2. Install the angle adjustment bracket 4: Secure the horizontal angle indexing flange 5 to the top of the support using anti-loosening screws. Then, place and fix the semi-circular pitch angle indexing plate 8 on the base plate. According to the approximate direction of the radar, rotate the horizontal angle indexing flange, select the corresponding hole position on the edge of the horizontal angle indexing flange, insert the anti-loosening screws and tighten them to complete the coarse adjustment and fixation of the horizontal azimuth angle.

[0118] 3. Install the pitch angle indexing plate and pitch adjustment: Install the pitch angle indexing plate on the horizontal angle indexing flange. Rotate the rotating shaft 7 to adjust the pitch angle of the radar corner reflector and the checkerboard vision plate. Then, insert the pin through the pin and fix it to the corner reflector support rod, with both ends of the pin fitting into the indexing holes 3 on the pitch angle indexing plate to position the pitch angle of the corner reflector support rod.

[0119] 4. Install the corner reflector and the chessboard vision panel: Install the radar corner reflector 1 to the end of the corner reflector support rod, and install the chessboard vision panel to the end of the support plate. The radar corner reflector constrains the rotational displacement of the chessboard vision panel.

[0120] 5. Final verification: Observe the echo signal of the target on the radar display screen, adjust the azimuth and elevation angles to maximize the radar echo intensity, and finally check and tighten all screws again.

[0121] Finally, any aspects not fully described in this invention utilize existing mature products and technologies.

[0122] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A railway slope monitoring method based on sensor fusion, characterized in that, Includes the following steps: Step 1: Plan and deploy fusion monitoring targets based on geological survey results; Step 2: Obtain the precise spatial coordinates and visual identification information of the fusion monitoring target, and generate QP values ​​based on the chessboard visual board on the target to form a mapping relationship table (ID, Xn, Yn, Zn, QP). Step 3: Sensor installation and debugging. Based on the on-site field of view, determine the optimal installation points for the radar and video surveillance; adjust the azimuth and elevation angles of the radar so that its beam can effectively cover all monitored targets; adjust the orientation and focal length of the camera so that it can clearly capture the target image. Step 4: Establish multi-sensor collaborative mapping relationships; Step 4.1: Calibrate Radar Image Pixel Coordinates: Read the precise coordinates of all targets and the radar's own pose parameters in the radar system, and perform coordinate transformation to establish a mapping table of "target ID - radar image pixel coordinates (ID, Radar_I, Radar_J)". Step 4.2: Establish a video sensor mapping table: Control the pan-tilt camera to point it at each target and the monitoring target of interest in turn; for each target, manually or automatically adjust the focus until the chessboard visual board image is clearest, and then record the azimuth angle of the pan-tilt at this time. Pitch angle and focal length parameters This is bound to the target and the target of interest ID, forming a mapping table of "target and target of interest ID - gimbal preset position parameters" (ID, , , Once completed, the camera captures a clear image of each target, which is then stored in the database as the initial background image Pic0. Step 4.3: Establish a radar-visual-target mapping table: Establish the mapping relationship between radar video targets using the unique ID of the target or object, forming a unique mapping relationship between each sensor based on the ID (ID). n Xn, Yn, Zn, QP n Radar_I n Radar_J n , n , n , n , , , (); where ID is the target code, X, Y, Z are the spatial coordinates of the target, QP is the target visual chessboard code, and Radar_I and Radar_J are the radar image metadata coordinates. For the target distance, The target azimuth angle, For the target scattering intensity, , , The preset azimuth, pitch, and focal length parameters of the camera relative to the target, where n represents the nth target; Step 5: Active radar monitoring to analyze the displacement and reflection intensity changes of the target in real time; In conjunction with periodic camera patrols and triggered checks, changes in the target's appearance are identified through image comparison. When either radar or camera detects an anomaly in the target, the other detection method is immediately triggered for collaborative verification. Finally, based on the results of the dual verification, a graded warning or alert is issued.

2. The railway slope monitoring method based on sensor fusion according to claim 1, characterized in that: Step 4.1 includes the following sub-steps: Step 4.1.1: Convert the world coordinate system to the radar's local rectangular coordinate system; First, to describe the positional offset of the target relative to the radar phase center, the translation vector T needs to be obtained: ； in, Geodetic coordinates of the target measured by RTK / total station; : Geodetic coordinates of the radar antenna phase center; Rotation matrix The parameter matrix is ​​used to rotate the world coordinate system (x, y, z) to the radar installation azimuth deflection direction; where a R is the radar's azimuth offset angle: ； Coordinate transformation result: The coordinates of the target point in the radar's local rectangular coordinate system are obtained. ; ； Step 4.1.2: Convert the radar's local rectangular coordinate system to the radar's polar coordinate system; Through geometric calculations, the slant range and azimuth angle used by the radar to sense the target are obtained; In the formula The straight-line spatial distance between the geodetic coordinates of the radar antenna phase center and the geodetic coordinates of the target center; Azimuth ; Step 4.1.3: Convert the radar polar coordinate system to radar cell index; Distance Gate Index : ； in, The minimum detection range set for the radar; Radar range resolution, ,in At the speed of light, For signal bandwidth; Azimuth Unit Index : ； in, The minimum azimuth angle for radar detection; Angular resolution of the radar; Calculate the range-angle unit that each target should theoretically appear in the radar image to establish a mapping table of "target ID - radar image unit coordinates (ID, Radar_I, Radar_J)"; This is used to calibrate radar data from relative measurements to absolute geographic coordinates, so that the displacement data monitored by the radar directly corresponds to the spatial displacement in the real world; the mapping relationship between radar image primitive coordinates (Radar_I, Radar_J) and target distance is also considered. Azimuth and scattering The value is unique.

3. The railway slope monitoring method based on sensor fusion according to claim 1, characterized in that: In Step 5, the edge processing unit processes the radar echo signal in real time, performing two analyses for each target pixel region in the mapping table: A. Displacement Analysis: Calculate the phase change of the current pixel and inversely determine the cumulative displacement of the target relative to its initial position. R; B. Reflection Intensity Analysis: Monitor the change in radar echo intensity of this pixel. ; Then the displacement is compared with the preset multi-level early warning thresholds. The RJ sensor is compared, including a noticeable value of 5mm, a warning value of 10mm, and an alarm value of 20mm, while simultaneously determining the reflection intensity. Does the system detect any mutations? If any condition exceeds the limit, the system immediately marks the target as "abnormal". When the radar detects an anomaly in the target, it triggers a verification process. At the same time, it sends a command to the camera through the mapping table, requiring it to immediately capture video image information of the abnormal target. Then, based on the video image detection AI algorithm, the target is visually identified, triggering a visual verification of the camera image.

4. The railway slope monitoring method based on sensor fusion according to claim 1, characterized in that: In Step 5, when the camera first detects an anomaly in the target, the radar verifies the anomaly. The radar receives the verification request from the video system and performs a rapid scan and data analysis on the designated target. If the radar verification result shows that the displacement and reflection intensity of the target are within limits, the system determines that this is a false alarm from the video and does not issue an alarm, but only records a "video detection change prompt" in the system log for management personnel to review. If the radar verification confirms that the displacement or intensity of the target has exceeded the limits, it indicates that the dual verification by radar and video is successful, and the risk of actual deformation of the slope is extremely high. The system immediately generates alarm information including alarm level, target location, displacement data, and on-site pictures, and issues it through various means such as sound and light, SMS, and platform push to remind relevant personnel to take emergency measures.

5. A system used in the railway slope monitoring method based on sensor fusion as described in claim 1, characterized in that: The system comprises a terminal layer for data interoperability, a network transmission layer, and an application service layer, wherein: The terminal layer includes slope radar, cameras, and edge computing terminals. The slope radar is used to collect slope surface displacement data; the cameras are used to collect slope image information; and the edge computing terminals are used to analyze, verify, and identify radar data and video data. Network transport layer: Used to transmit data collected by the terminal layer; Application service layer: This includes a radar video surveillance and early warning platform, which is used to receive, process, and display monitoring data and generate early warning information.

6. A target used in the railway slope monitoring method based on sensor fusion as described in claim 1, characterized in that: It includes a support, an angle adjustment frame, a radar corner reflector, and a checkerboard vision panel. The top of the support rotates to support the angle adjustment frame on a horizontal plane. The angle adjustment frame supports the radar corner reflector and the checkerboard vision panel by adjusting the pitch angle. The checkerboard vision panel has multiple squares arranged in alternating black and white colors.

7. The target used in the railway slope monitoring method based on sensor fusion according to claim 6, characterized in that: The bottom of the angle adjustment frame is fixed with a horizontal angle indexing flange, the middle of the angle adjustment frame is horizontally inserted with a rotating shaft, and the top of the angle adjustment frame is made into a pitch angle indexing plate with a top arc surface; the edge of the pitch angle indexing plate is provided with multiple horizontally penetrating indexing holes. The horizontal angle-dividing flange is adjusted horizontally on the top plane of the support and bolted together. The radar corner reflector and the chessboard vision plate are rotatably mounted on a rotating shaft at one end, with the other end of the radar corner reflector pinned to the indexing hole, and the other end of the chessboard vision plate overlapping the radar corner reflector for pitch angle positioning constraint.

Images