A train inspection robot autonomous inspection method, system and storage medium

Abstract

The present invention discloses an autonomous inspection method, system and storage medium for a train inspection robot, which belongs to the technical field of inspection robots, and includes: respectively receiving a first inspection instruction, a second inspection instruction and inspection information sent by a host computer, the inspection information at least including the parking lane information of the train to be inspected, a first inspection list, a second inspection list and train formation information; based on the first inspection instruction, loading a 3D map and a first inspection list outside the corresponding parking lane according to the parking lane information, and inspecting both sides of the train to be inspected; based on the second inspection instruction, loading a chassis map and a second inspection list of the train to be inspected, and inspecting the chassis of the train to be inspected. The present invention realizes autonomous navigation and positioning in a narrow and similar environment in a train parking lane and in an open and similar environment outside a train parking lane based on multi-sensor perception technology, and implements defect detection.

Description

Technical Field

[0001] The present invention relates to the technical field of inspection robots, and in particular to an autonomous inspection method, system and storage medium for a train inspection robot. Background Art

[0002] Urban rail transit, as a means of transportation with high capacity, high speed, low energy consumption, minimal pollution, and high reliability, is a public transportation system prioritized by countries around the world. With the massive investment and use of subways, routine inspection and maintenance are essential.

[0003] Subway vehicle inspections are a daily necessity for subway operations, particularly requiring detailed chassis inspections to promptly identify safety hazards. Subway inspections are challenging and currently performed manually. Replacing manual labor with machines is a pressing need for railway authorities. However, due to the high technical complexity and specialized testing requirements, mature products are not readily available on the market.

[0004] Furthermore, due to differences in the capabilities and overall quality of maintenance personnel, the quality of maintenance also varies. To improve inspection quality and efficiency, the government has recently introduced modern track maintenance equipment and systems to assist with urban rail transit train maintenance. However, the automation, informationization, and intelligence levels of maintenance equipment and systems within urban rail transit sections are low, resulting in a poorly integrated and interconnected urban rail transit train inspection process. Maintenance personnel are still required to participate in most of the maintenance process, preventing the implementation of truly intelligent maintenance.

[0005] Train inspection robots will gradually replace manual labor due to their advantages such as fast maintenance methods, standardized maintenance, and the ability to adjust maintenance items at any time. Among them, the core difficulties of autonomous robot inspections are navigation and positioning technology and AI defect detection technology.

[0006] There are two main navigation technology solutions in tunnels:

[0007] 1) Laying tracks or adding auxiliary markers in the tunnel. The disadvantage of this method is that it requires construction and renovation of the tunnel, which affects normal maintenance.

[0008] 2) Mapping the tunnels. This method suffers from a drawback in that the location where the train stops has centimeter-level errors, which affects the defect detection and recognition rate.

[0009] The environment outside the tunnel is relatively similar, the route is long, and there are also situations where trains entering and exiting cause changes in the environment. Currently, there are two main navigation technology solutions outside the tunnel:

[0010] 1) Laying auxiliary markers such as magnetic strips, magnetic nails, and reflective panels on the ground outside the tunnel. This has the disadvantage that it requires construction, and routine maintenance by maintenance personnel may cause the auxiliary markers to fall off or be covered, affecting the normal inspection of the robot.

[0011] 2) Laser SLAM navigation. This method has the disadvantage that, because the scene length is longer than the entire train and the environment is constantly changing, the laser's sensing distance is short, making it impossible to achieve stable mapping and navigation.

[0012] Defect identification can be divided into nine categories, namely crack detection, anti-loosening mark detection, brake wear detection, tread detection, temperature detection, relay temperature measuring sticker reading, oil level reading, surface anomaly and defect detection, and surface oil leakage detection. Since it involves multi-sensor collaboration and data fusion analysis, existing technologies fully cover the above detection. Summary of the Invention

[0013] The purpose of the present invention is to provide an autonomous inspection method and system for a train inspection robot, aiming to achieve autonomous navigation and positioning in a narrow and similar environment inside a train parking aisle and in an open and similar environment outside a train parking aisle.

[0014] To achieve the above objectives, the present invention provides an autonomous inspection method for a train inspection robot, comprising:

[0015] Respectively receiving a first inspection instruction, a second inspection instruction, and inspection information sent by the host computer, wherein the inspection information at least includes the parking lane information of the train to be inspected, the first inspection list, the second inspection list, and train formation information;

[0016] Based on the first inspection instruction, loading a 3D map outside the corresponding parking lane and a first inspection list according to the parking lane information, and inspecting both sides of the train to be inspected;

[0017] Based on the second inspection instruction, the chassis map and the second inspection list of the train to be inspected are loaded, and the chassis of the train to be inspected is inspected.

[0018] Furthermore, based on the first inspection instruction, loading a 3D map outside the corresponding parking lane and a first inspection list according to the parking lane information, and inspecting both sides of the train to be inspected includes:

[0019] Based on the laser sensor and the first depth camera carried by the train inspection robot, laser data and perception data are respectively acquired;

[0020] Fusing the laser data and the perception data, and performing positioning and navigation on the 3D map based on the fusion result;

[0021] During the navigation process, a task scheduling algorithm is executed to control the industrial camera carried by the train inspection robot to take pictures of the inspection area of ​​the train to be inspected to obtain pictures to be inspected;

[0022] The image to be detected is analyzed using a deep learning-based visual detection algorithm, and the analysis results are uploaded to the host computer.

[0023] Furthermore, the loading of the chassis map and the second inspection list of the train to be inspected based on the second inspection instruction, and inspecting the chassis of the train to be inspected, includes:

[0024] Controlling the elevation angle of the laser sensor carried by the train inspection robot to 45 degrees, and realizing positioning based on the train chassis and patrol navigation of the center of the parking lane on the chassis map based on the data collected by the laser sensor and depth camera carried by the train inspection robot;

[0025] During the navigation process, the laser sensor and the industrial camera carried by the train inspection robot are controlled to photograph the inspection area of ​​the train to be inspected, and 3D laser data and 2D data are obtained respectively;

[0026] The 3D laser data and 2D data are analyzed, and the analysis results are uploaded to the host computer.

[0027] Furthermore, the control of the elevation angle of the laser sensor carried by the train inspection robot to 45 degrees and the implementation of positioning based on the train chassis and parking lane center patrol navigation on the chassis map based on data collected by the laser sensor and depth camera carried by the train inspection robot include:

[0028] Control the laser sensor and the first depth camera carried by the train inspection robot to have an elevation angle of 45 degrees, and collect chassis data of the train to be inspected through the laser sensor and the first depth camera;

[0029] Fuse the collected chassis data to achieve train chassis positioning;

[0030] The second depth camera and the third depth camera carried by the train inspection robot are used to extract the walls on both sides of the parking lane, obtain the centerline position of the parking lane, and realize parking lane center line patrol navigation. The second depth camera and the third depth camera are front and rear horizontal depth cameras respectively.

[0031] Furthermore, during the navigation process, the process of controlling the camera carried by the train inspection robot to photograph the area to be inspected of the train to be inspected includes:

[0032] Get the control pose (x1, y1, z1, r1, o1, p1) of the manipulator arm for capturing an image at a certain inspection point, where x1, y1, z1 are the three-dimensional coordinates of the end of the manipulator arm relative to the base, and r1, o1, p1 are the rotation, pitch, and roll angles of the end of the manipulator arm relative to the base, respectively.

[0033] The positioning algorithm is used to output the three-dimensional positioning (x2, y2, z2, r2, o2, p2). The coordinates of the inspection point are (x3, y3, z3, r3, o3, p3). The coordinates of the robot arm controlled shooting are calculated as (x1+x2-x3, y1+y2-y3, z1+z2-z3, r1+r2-r3, o1+o2-o3, p1+p2-p3). x2, y2, z2 represent the three-dimensional coordinates of the robot in the train chassis map during autonomous inspection, r2, o2, p2 are the rotation, pitch and roll angles of the robot in the train chassis map during autonomous inspection, x3, y3, z3 represent the three-dimensional coordinates of the robot in the train chassis map during teaching, r3, o3, p3 are the rotation, pitch and roll angles of the robot in the train chassis map during teaching;

[0034] The industrial camera is controlled to shoot according to the coordinates controlled by the robotic arm.

[0035] Furthermore, during the navigation process, the process of controlling the laser sensor to photograph the area to be inspected of the train to be inspected includes:

[0036] According to the position of the area to be inspected, the robot arm of the train inspection robot is controlled to move to the inspection preparation points, which are four points taught in advance and correspond to the four quadrants at the top of the train inspection robot;

[0037] Calculating the detection starting point and the detection ending point according to the position of the area to be detected;

[0038] The laser sensor is controlled to start the line scanning laser at the detection starting point, and ends the line scanning after moving horizontally to the detection ending point.

[0039] Furthermore, the analysis of the 3D laser data and 2D data includes at least one of crack detection, anti-loosening mark detection, brake wear detection, tread detection, temperature detection, relay temperature measuring sticker reading recognition, oil level recognition, surface anomaly and defect detection, and surface oil leakage detection.

[0040] Furthermore, after the inspection of the train to be inspected is completed, the method further includes:

[0041] The train inspection robot automatically returns to the charging room or performs the inspection task of the next train.

[0042] In addition, to achieve the above objectives, the present invention also provides an autonomous inspection system for a train inspection robot, comprising: a train inspection robot and a host computer, wherein the train inspection robot is equipped with a laser sensor, a depth camera, an industrial camera, and a control device, wherein the control device comprises:

[0043] An information receiving module is used to receive the first inspection instruction, the second inspection instruction and the inspection information sent by the host computer respectively, wherein the inspection information at least includes the parking lane information of the train to be inspected, the first inspection list, the second inspection list and the train formation information;

[0044] A first inspection module is configured to load a 3D map outside the corresponding parking lane and a first inspection list according to the parking lane information based on the first inspection instruction, and inspect both sides of the train to be inspected;

[0045] The second inspection module is used to load the chassis map and the second inspection list of the train to be inspected based on the second inspection instruction, and inspect the chassis of the train to be inspected.

[0046] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium on which computer-readable instructions are stored. The computer-readable instructions can be executed by a processor to implement the above-mentioned autonomous inspection method of the train inspection robot.

[0047] Compared with the existing technology, the present invention has the following technical effects: In the present invention, after the train enters the maintenance track, the host computer main control sends inspection instructions and inspection information to the two train inspection robots. After the robot group loads the map and preset path according to the lane information, it drives to the inspection starting point on both sides of the train side track to start the daily inspection process of the target train. For the exterior of the train, the robot group performs non-stop inspection on both sides, positioning and navigating based on the horizontal front 3D laser and depth camera, and controlling the pan-tilt angle so that the high-definition industrial camera can capture the area to be inspected and conduct real-time analysis. After the exterior is completed, the robot group drives into the tunnel via a slope one after the other and controls the 3D laser to tilt 45 degrees horizontally. Using the train chassis map and based on the 3D laser tilted 45 degrees horizontally, the depth camera tilted 45 degrees horizontally, and the horizontal front and rear depth cameras for positioning and navigation, the robot group divides the work and operates in coordination with the centerline of the train chassis as the dividing line. At each inspection point, the robot arm controls the line scanning laser and 2D high-definition camera at the end to collaboratively acquire and analyze data from the area to be inspected, thereby completing the train inspection task. The present invention builds an autonomous navigation system, an AI defect detection system, and an autonomous inspection system, integrates various types of detection and inspection data, completes the aggregation of perception resources, and combines it with the application needs of actual scenarios such as subway train chassis to solve the technical difficulties of train inspection robots completing related tasks in complex scenarios such as subway train chassis. It can improve the knowability and controllability of rail transit inspection content, realize the automation, intelligence, and convenience of daily inspection of rail transit trains, and solve the problems of high labor intensity, high density, and low accuracy of rail transit train inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings:

[0049] Figure 1 This is a flow chart of an autonomous inspection method of a train inspection robot according to the present invention;

[0050] Figure 2 2. It is a schematic diagram of line scanning of the laser sensor of the present invention;

[0051] Figure 3 This is a schematic diagram of the robot group's cooperation inside and outside the parking lane of the present invention. DETAILED DESCRIPTION

[0052] To further illustrate the features of the present invention, please refer to the following detailed description of the present invention and the accompanying drawings. The accompanying drawings are for reference and illustration purposes only and are not intended to limit the scope of protection of the present invention.

[0053] like Figure 1 As shown, this embodiment discloses an autonomous inspection method of a train inspection robot, comprising the following steps S1 to S3:

[0054] S1. Respectively receive a first inspection instruction, a second inspection instruction, and inspection information sent by a host computer, wherein the inspection information includes at least the parking lane information of the train to be inspected, a first inspection list, a second inspection list, and train formation information;

[0055] S2. Based on the first inspection instruction, load a 3D map outside the corresponding parking lane and a first inspection list according to the parking lane information, and inspect both sides of the train to be inspected;

[0056] S3. Based on the second inspection instruction, load the chassis map and the second inspection list of the train to be inspected, and inspect the chassis of the train to be inspected.

[0057] It should be noted that, in this embodiment, after the train to be inspected enters the maintenance track, the host computer main control sends the first inspection instruction, the second inspection instruction and inspection information to the two idle robots. After the robot group loads the map and preset path according to the lane information, it drives to the inspection starting point on both sides of the train side track to start the daily maintenance process of the target train, including inspection of the outside and bottom of the train, and uploads the inspection results to the host computer.

[0058] It should be noted that the robot is equipped with a laser sensor, a first depth camera, a second depth camera, a third depth camera and a high-definition industrial camera, wherein the angles of the laser sensor and the first depth camera are adjustable, and the second depth camera and the third depth camera are arranged horizontally.

[0059] As a further preferred technical solution, step S1: based on the first inspection instruction, loading the 3D map outside the corresponding parking lane and the first inspection list according to the parking lane information, and inspecting both sides of the train to be inspected includes the following steps S11 to S14:

[0060] S11. Acquire laser data and perception data based on a laser sensor and a first depth camera carried by the train inspection robot;

[0061] S12, fusing the laser data and the perception data, and performing positioning navigation on the 3D map according to the fusion result;

[0062] S13. During the navigation process, executing a task scheduling algorithm to control the industrial camera carried by the train inspection robot to photograph the inspection areas on both sides of the train to be inspected, thereby obtaining images to be inspected;

[0063] It should be noted that the task scheduling algorithm is executed to assign tasks to the two robots. For example, one robot inspects the left side of the train and the other inspects the right side of the train. After inspecting both sides, they line up at the entrance of the maintenance road and enter the maintenance road in turn for inspection.

[0064] S14: Analyze the image to be detected using a deep learning-based visual detection algorithm, such as YOLOv5, and upload the analysis results to the host computer.

[0065] As a further preferred technical solution, the step S2: loading the chassis map and the second inspection list of the train to be inspected based on the second inspection instruction, and inspecting the chassis of the train to be inspected, includes the following steps S21 to S23:

[0066] S21. Control the elevation angle of the laser sensor carried by the train inspection robot to 45 degrees, and implement positioning based on the train chassis and parking lane center patrol navigation on the chassis map based on data collected by the laser sensor and depth camera carried by the train inspection robot;

[0067] S22. During the navigation process, control the laser sensor and the industrial camera carried by the train inspection robot to photograph the inspection area of ​​the train to be inspected, and obtain 3D laser data and 2D data respectively;

[0068] S23, analyzing the 3D laser data and the 2D data, and uploading the analysis results to the host computer.

[0069] As a further preferred technical solution, step S21: controlling the elevation angle of the laser sensor carried by the train inspection robot to 45 degrees, and implementing positioning based on the train chassis and parking lane center patrol navigation on the chassis map based on data collected by the laser sensor and depth camera carried by the train inspection robot, specifically includes:

[0070] Control the laser sensor and the first depth camera carried by the train inspection robot to have an elevation angle of 45 degrees, and collect chassis data of the train to be inspected through the laser sensor and the first depth camera;

[0071] Fuse the collected chassis data to achieve train chassis positioning;

[0072] The second depth camera and the third depth camera carried by the train inspection robot are used to extract the walls on both sides of the parking lane, obtain the centerline position of the parking lane, and realize parking lane center line patrol navigation. The second depth camera and the third depth camera are front and rear horizontal depth cameras respectively.

[0073] It should be noted that the positioning algorithm in this embodiment uses a 3D laser and depth camera with a 45-degree elevation angle. By fusing homogeneous chassis perception data, it enables a more similar description of the chassis of different trains of the same model, thereby achieving stable positioning. The navigation algorithm in the parking lane uses front and rear horizontal depth cameras to extract the walls on both sides of the parking lane and determine the centerline position.

[0074] As a further preferred technical solution, during the navigation process, the process of controlling the camera carried by the train inspection robot to photograph the area to be inspected of the train to be inspected includes:

[0075] Get the control pose (x1, y1, z1, r1, o1, p1) of the manipulator arm for capturing an image at a certain inspection point, where x1, y1, z1 are the three-dimensional coordinates of the end of the manipulator arm relative to the base, and r1, o1, p1 are the rotation, pitch, and roll angles of the end of the manipulator arm relative to the base, respectively.

[0076] The positioning algorithm is used to output the three-dimensional positioning (x2, y2, z2, r2, o2, p2). The coordinates of the inspection point are (x3, y3, z3, r3, o3, p3). The coordinates of the robot arm controlled shooting are calculated as (x1+x2-x3, y1+y2-y3, z1+z2-z3, r1+r2-r3, o1+o2-o3, p1+p2-p3). x2, y2, z2 represent the three-dimensional coordinates of the robot in the train chassis map during autonomous inspection, r2, o2, p2 are the rotation, pitch and roll angles of the robot in the train chassis map during autonomous inspection, x3, y3, z3 represent the three-dimensional coordinates of the robot in the train chassis map during teaching, r3, o3, p3 are the rotation, pitch and roll angles of the robot in the train chassis map during teaching;

[0077] The industrial camera is controlled to shoot according to the coordinates controlled by the robotic arm.

[0078] As a further preferred technical solution, Figure 2 As shown, during the navigation process, the process of controlling the laser sensor to photograph the area to be inspected of the train to be inspected includes:

[0079] According to the position of the area to be inspected, the robot arm of the train inspection robot is controlled to move to the inspection preparation points, which are four points taught in advance and correspond to the four quadrants at the top of the train inspection robot;

[0080] Calculating the detection starting point and the detection ending point according to the position of the area to be detected;

[0081] The laser sensor is controlled to start the line scanning laser at the detection starting point, and ends the line scanning after moving horizontally to the detection ending point.

[0082] As a further preferred technical solution, the 3D laser data and 2D data are analyzed, including at least one of crack detection, anti-loosening mark detection, brake wear detection, tread detection, temperature detection, relay temperature measuring sticker reading recognition, oil level recognition, surface anomaly and defect detection, and surface oil leakage detection.

[0083] It should be noted that this embodiment performs analysis based on 2D visual data and 3D laser data, and implements nine categories of defect detection, specifically:

[0084] 1) For crack detection, 2D high-definition camera data and 3D laser data are integrated, and deep learning algorithms are used to identify and simultaneously detect crack depth;

[0085] 2) For anti-loosening mark detection, identify the red line in the 2D high-definition image and use 3D laser data registration to calculate the horizontal and deflection errors;

[0086] 3) For brake wear detection, only 3D laser data is used to identify the brake pad and brake pad retainer and calculate the height difference between the two;

[0087] 4) For tread inspection, only 3D laser data is used, scratches are detected by depth threshold, and wear is detected based on the height difference between scratches and the rim thickness;

[0088] 5) For temperature detection, only 2D infrared data is used. An infrared thermal imager is used to capture the area to be detected. The device to be measured is extracted using a contour detection algorithm. The temperature of the area within the contour is measured, and the maximum temperature, average temperature, and minimum temperature are output.

[0089] 6) For relay temperature measurement patch reading recognition, only 2D high-definition data is used to detect the position of the meter bar and output the result as a percentage based on the calibration data;

[0090] 7) For oil level identification, only 2D high-definition data is used to detect the oil level line and calculate its position, and the result is output as a percentage based on the calibration data;

[0091] 8) For surface anomaly and defect detection, 2D high-definition and 3D laser data are integrated and a deep learning-based target detection algorithm is used for identification;

[0092] 9) For surface oil leak detection, only 2D high-definition data is used, and a deep learning-based target detection algorithm is used for identification.

[0093] As a further preferred technical solution, after the inspection of the train to be inspected is completed, the following step S4 is further included:

[0094] S4. The train inspection robot automatically returns to the charging room or performs the inspection task of the next train.

[0095] It should be noted that after the train enters the maintenance track, the host computer sends inspection instructions and related information to two idle robots. The robot team, after loading a map and pre-set route based on the lane information, then drives to the inspection starting points on both sides of the train's side track to begin the routine inspection process for the target train. For the train's exterior, the robot team performs non-stop inspections on both sides, using a horizontal front-mounted 3D laser and depth camera for positioning and navigation. The robot also controls the pan / tilt angle to enable the high-definition camera to capture and analyze the area to be inspected in real time. After completing the exterior inspection, the robot team, one in front and one behind, drives down the ramp into the tunnel and tilts the 3D laser upward 45 degrees from the horizontal. Using the train chassis map and using the 45-degree 3D laser, 45-degree depth camera, and horizontal front and rear depth cameras for positioning and navigation, the robot team divides the work and operates collaboratively, with the train chassis centerline as the dividing line. At each inspection point, the robot arm controls the line-scanning laser at the end and the 2D high-definition camera to collaboratively acquire and analyze data from the area to be inspected. After the inspection is completed, the robot team automatically returns to the charging room or proceeds to inspect the next train.

[0096] The present invention realizes autonomous navigation and positioning in narrow and similar environments inside train parking lanes and in open and similar environments outside train parking lanes based on multi-sensor perception technology, and realizes nine major categories of defect detection.

[0097] like Figure 3 As shown, the autonomous inspection steps of the train inspection robot in this embodiment are as follows:

[0098] The robot is mainly equipped with a 2D high-definition infrared dual-camera gimbal, a pitchable 3D laser module, three depth cameras (one each for the front, back, and 45-degree elevation), a 6-degree-of-freedom mechanism, a line-scan laser, a 2D high-definition industrial camera and other equipment.

[0099] (1) Control robot No. 1 to drive from the charging room to the outside of the parking lane and circle the train to build a real-time 3D map.

[0100] (2) After the map outside the parking lane is constructed, the inspection path and pan-tilt angle are set, and the inspection regulations are compared so that the 2D high-definition camera in the pan-tilt can capture the area to be inspected.

[0101] (3) After the inspection and maintenance regulations are implemented, the inspection task will be issued to Robot No. 1 for preliminary inspection testing and optimization. After the requirements are met, the data of Robot No. 1 will be imported into Robot No. 2 for inspection verification and optimization until the inspection requirements are met.

[0102] (4) Control the two robots to enter the inspection lane separately, and use the centerline of the train chassis as the dividing line. Refer to the inspection regulations, control the two robots' robotic arms, end-point 2D high-definition cameras, and end-point line scanning lasers to inspect the area to be inspected on the chassis, and save the robotic arm control data, high-definition images, 3D point cloud data, inspection type, etc.

[0103] (5) After the chassis is fully implemented, the two robots are assigned chassis inspection tasks for testing and optimization until the inspection requirements are met.

[0104] (6) A map switching point is set on the map outside the parking lane so that the robot can switch to the train chassis map before entering the parking lane, thereby being able to autonomously drive from the ramp to the parking lane.

[0105] (7) Issue an overall inspection task, so that the robot group starts from the charging pile and works together to complete the autonomous inspection outside and inside the parking lane in turn.

[0106] This embodiment also provides an autonomous inspection system for a train inspection robot, comprising: a train inspection robot and a host computer. The train inspection robot is equipped with a laser sensor, a depth camera, an industrial camera, and a control device. The control device includes:

[0107] An information receiving module is used to receive the first inspection instruction, the second inspection instruction and the inspection information sent by the host computer respectively, wherein the inspection information at least includes the parking lane information of the train to be inspected, the first inspection list, the second inspection list and the train formation information;

[0108] A first inspection module is configured to load a 3D map outside the corresponding parking lane and a first inspection list according to the parking lane information based on the first inspection instruction, and inspect both sides of the train to be inspected;

[0109] The second inspection module is used to load the chassis map and the second inspection list of the train to be inspected based on the second inspection instruction, and inspect the chassis of the train to be inspected.

[0110] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of the embodiment scheme, and no limitation is made here.

[0111] In addition, for technical details not fully described in this embodiment, reference can be made to the writing method provided in any embodiment of the present invention, and no further details will be given here.

[0112] In addition, an embodiment of the present invention further proposes a computer-readable storage medium on which computer-readable instructions are stored. The computer-readable instructions can be executed by a processor to implement the autonomous inspection method of the train inspection robot as described above.

[0113] Since the storage medium adopts all the technical solutions of all the above embodiments, it has at least all the beneficial effects brought by the technical solutions of the above embodiments, which will not be described one by one here.

[0114] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, or of course by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as a read-only memory (ROM) / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a writing device, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0115] 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 principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A train inspection robot autonomous inspection method, characterized in that: include: S1, respectively receiving the first inspection instruction, the second inspection instruction and the inspection information sent by the host computer, wherein the inspection information at least includes the parking lane information of the train to be inspected, the first inspection list, the second inspection list and the train formation information; S2. Based on the first inspection instruction, load the 3D map outside the corresponding parking lane and the first inspection list according to the parking lane information, and inspect both sides of the train to be inspected; S3. Based on the second inspection instruction, load the chassis map and the second inspection list of the train to be inspected, and inspect the chassis of the train to be inspected; Wherein step S2 includes steps S21 to S23: S21, controlling the elevation angle of the laser sensor carried by the train inspection robot to be 45 degrees, and realizing the positioning based on the train chassis and the patrol line navigation in the center of the parking lane on the chassis map according to the data collected by the laser sensor and the depth camera carried by the train inspection robot; S22. During the navigation process, control the laser sensor and the industrial camera carried by the train inspection robot to photograph the inspection area of ​​the train to be inspected, and obtain 3D laser data and 2D data respectively; S23, analyzing the 3D laser data and the 2D data, and uploading the analysis results to the host computer; In step S22, during the navigation process, the process of controlling the camera carried by the train inspection robot to photograph the inspection area of ​​the train to be inspected includes: Get the robot control posture (x1, y1, z1, r1, o1, p1) of the image taken at a certain inspection point, where x1, y1, z1 are the three-dimensional coordinates of the end of the robot relative to the base, and r1, o1, p1 are the rotation, pitch and roll angles of the end of the robot relative to the base; The positioning algorithm is used to output the three-dimensional positioning (x2, y2, z2, r2, o2, p2). The coordinates of the inspection point are (x3, y3, z3, r3, o3, p3). The coordinates of the robot arm control shooting are calculated as (x1+x2-x3, y1+y2-y3, z1+z2-z3, r1+r2-r3, o1+o2-o3, p1+p2-p3). x2, y2, z2 represent the three-dimensional coordinates of the robot in the train chassis map during autonomous inspection, r2, o2, p2 are the rotation, pitch and roll angles of the robot in the train chassis map during autonomous inspection, x3, y3, z3 represent the three-dimensional coordinates of the robot in the train chassis map during teaching, r3, o3, p3 are the rotation, pitch and roll angles of the robot in the train chassis map during teaching; The industrial camera is controlled to shoot according to the coordinates controlled by the robotic arm.

2. The autonomous inspection method of a train inspection robot according to claim 1, characterized in that: The step of loading a 3D map and a first inspection list outside the corresponding parking lane according to the parking lane information based on the first inspection instruction and inspecting both sides of the train to be inspected includes: Based on the laser sensor and the first depth camera carried by the train inspection robot, laser data and perception data are respectively acquired; Fusing the laser data and the perception data, and performing positioning and navigation on the 3D map according to the fusion result; During the navigation process, a task scheduling algorithm is executed to control the industrial camera carried by the train inspection robot to take pictures of the inspection area of ​​the train to be inspected to obtain pictures to be inspected; The image to be detected is analyzed using a deep learning-based visual detection algorithm, and the analysis results are uploaded to the host computer.

3. The autonomous inspection method of a train inspection robot according to claim 1, characterized in that: The control of the elevation angle of the laser sensor carried by the train inspection robot to 45 degrees and the realization of positioning based on the train chassis and patrol line navigation in the center of the parking lane on the chassis map according to the data collected by the laser sensor and the depth camera carried by the train inspection robot include: Control the laser sensor and the first depth camera carried by the train inspection robot to have an elevation angle of 45 degrees, and collect chassis data of the train to be inspected through the laser sensor and the first depth camera; The collected chassis data is integrated to realize the train chassis positioning; The second depth camera and the third depth camera carried by the train inspection robot are used to extract the walls on both sides of the parking lane, obtain the centerline position of the parking lane, and realize the parking lane center line patrol navigation. The second depth camera and the third depth camera are front and rear horizontal depth cameras respectively.

4. The autonomous inspection method of a train inspection robot according to claim 1, characterized in that: During the navigation process, the process of controlling the laser sensor to photograph the area to be inspected of the train to be inspected includes: According to the position of the area to be inspected, the mechanical arm of the train inspection robot is controlled to move to the inspection preparation points, where the inspection preparation points are four points taught in advance, corresponding to the four quadrants at the top of the train inspection robot; Calculate the detection start point and the detection end point according to the position of the area to be detected; The laser sensor is controlled to start the line scanning laser at the detection starting point, and ends the line scanning after moving horizontally to the detection ending point.

5. The autonomous inspection method of a train inspection robot according to claim 1, characterized in that: The analysis of the 3D laser data and the 2D data includes at least one of crack detection, anti-loosening mark detection, brake wear detection, tread detection, temperature detection, relay temperature measuring sticker reading recognition, oil level recognition, surface anomaly and defect detection, and surface oil leakage detection.

6. The autonomous inspection method of a train inspection robot according to claim 1, characterized in that: After the inspection of the train to be inspected is completed, the method further includes: The train inspection robot automatically returns to the charging room or performs the inspection task of the next train.

7. A train inspection robot autonomous inspection system, characterized in that: include: The train inspection robot and the host computer are equipped with a laser sensor, a depth camera, an industrial camera and a control device. The control device includes: An information receiving module, used to respectively receive a first inspection instruction, a second inspection instruction and inspection information sent by a host computer, wherein the inspection information at least includes the parking lane information of the train to be inspected, a first inspection list, a second inspection list and train marshaling information; A first inspection module, configured to load a 3D map outside the corresponding parking lane and a first inspection list according to the parking lane information based on the first inspection instruction, and inspect both sides of the train to be inspected; A second inspection module is used to load the chassis map and the second inspection list of the train to be inspected based on the second inspection instruction, and inspect the chassis of the train to be inspected; The first inspection module includes steps S21 to S23: S21, controlling the elevation angle of the laser sensor carried by the train inspection robot to be 45 degrees, and realizing the positioning based on the train chassis and the patrol line navigation in the center of the parking lane on the chassis map according to the data collected by the laser sensor and the depth camera carried by the train inspection robot; S22. During the navigation process, control the laser sensor and the industrial camera carried by the train inspection robot to photograph the inspection area of ​​the train to be inspected, and obtain 3D laser data and 2D data respectively; S23, analyzing the 3D laser data and the 2D data, and uploading the analysis results to the host computer; In step S22, during the navigation process, the process of controlling the camera carried by the train inspection robot to photograph the inspection area of ​​the train to be inspected includes: Get the robot control posture (x1, y1, z1, r1, o1, p1) of the image taken at a certain inspection point, where x1, y1, z1 are the three-dimensional coordinates of the end of the robot relative to the base, and r1, o1, p1 are the rotation, pitch and roll angles of the end of the robot relative to the base; The positioning algorithm is used to output the three-dimensional positioning (x2, y2, z2, r2, o2, p2). The coordinates of the inspection point are (x3, y3, z3, r3, o3, p3). The coordinates of the robot arm control shooting are calculated as (x1+x2-x3, y1+y2-y3, z1+z2-z3, r1+r2-r3, o1+o2-o3, p1+p2-p3). x2, y2, z2 represent the three-dimensional coordinates of the robot in the train chassis map during autonomous inspection, r2, o2, p2 are the rotation, pitch and roll angles of the robot in the train chassis map during autonomous inspection, x3, y3, z3 represent the three-dimensional coordinates of the robot in the train chassis map during teaching, r3, o3, p3 are the rotation, pitch and roll angles of the robot in the train chassis map during teaching; The industrial camera is controlled to shoot according to the coordinates controlled by the robotic arm.

8. A computer-readable storage medium, characterized in that: Computer-readable instructions are stored thereon, and the computer-readable instructions can be executed by a processor to implement the autonomous inspection method of the train inspection robot as described in any one of claims 1-6.

Images