Railway contact net mechanical deicing method and system and storage medium

Abstract

The invention discloses a mechanical deicing method and system for a railway overhead line system and a storage medium. The method comprises the steps that a non-contact measurement system conducts preparatory work; carrying out a calibration method based on plane constraint, and enabling the point cloud data collected by the laser radar and the two-dimensional image data collected by the camera to be correlated through a parameter conversion mode; the method comprises the following steps of: basically limiting a geometric parameter measurement original value of the overhead line system according to a railway overhead line system design standard, and performing data screening to obtain effective measurement data; based on a deep learning model and a laser radar, positioning and distance measurement of the contact network, a contact network dropper and a contact network positioner are completed; the automatic obstacle avoidance of the mechanical deicing device is realized; the safety range of the robot and the safety range of a deicing tool at the tail end of the robot are limited; and the robot end tool position and posture are adjusted for deicing. According to the device, ice covering the contact net is automatically removed, the limitation of manual deicing is avoided, and the deicing efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of railway contact network maintenance, and more specifically, relates to a mechanical deicing method, system and storage medium for railway contact network. Background Art

[0002] With the widespread laying and operation of electrified railways around the world, the contact network, as a key component of its power supply system, undertakes the important mission of transmitting electric energy to electric locomotives. However, since the contact network is mostly exposed to the outdoor environment, it is very easy to be covered with ice in high-altitude cold areas or under certain severe climatic conditions, such as low temperatures, freezing rain, snowfall and other weather conditions in winter.

[0003] This icing phenomenon has brought many severe challenges to the normal operation of electrified railways. From the perspective of the scope of the affected areas, it often involves long sections along the railway. Once the icing is serious, it will cause a large-scale power outage, which will paralyze rail transportation. In terms of deicing operations, the traditional main deicing method relies on workers to manually knock on the contact wire to remove the ice. However, there are many disadvantages that are difficult to overcome in manual deicing. First, its operating efficiency is extremely low, the speed and coverage of manual operation are limited, and it is difficult to effectively deal with a large area of ​​ice-covered contact wire in a short time, which seriously affects the normal operation time and transportation efficiency of the railway. Secondly, the safety risk is extremely high. Workers need to work in a high-altitude environment above the railway, facing a variety of dangerous hazards such as falling and electric shock, which poses a serious threat to the life safety of workers. Furthermore, manual deicing requires a lot of manpower, material and financial resources, including the labor costs of many workers, the purchase costs of protective equipment, the consumption of deicing tools, and the indirect economic losses caused by railway operation delays caused by operations.

[0004] In addition, the existing deicing technology is unable to cope with the complex structure and surrounding components of the contact network. There are many components around the contact network, such as suspension strings and positioners. During the deicing process, if these components cannot be accurately avoided, not only may these auxiliary components of the contact network be damaged, affecting their normal function and service life, but also the deicing device itself may be damaged due to collision, further increasing maintenance costs and operational difficulties.

[0005] In summary, there is an urgent need to develop a new and efficient railway contact network de-icing device and method to overcome the defects of the existing technology, achieve effective removal of ice from the contact network, while ensuring the safety of the contact network and its surrounding components, improve the automation level and efficiency of de-icing operations, reduce manpower and material costs, and ensure the safe, stable and efficient operation of electrified railways in cold climate conditions. Summary of the invention

[0006] In order to solve the above problems, it is necessary to study a method, system and storage medium for mechanical deicing of railway contact network, which can effectively remove thicker ice on the contact network. At the same time, the deicing device does not damage the contact network wires and can actively avoid components around the contact network.

[0007] In view of the above defects or improvement needs of the prior art, according to a first aspect of the present invention, a method for mechanical deicing of a railway contact network is provided, comprising:

[0008] S1.Prepare the non-contact measurement system;

[0009] S2. Perform a calibration method based on plane constraints to associate the point cloud data collected by the lidar and the two-dimensional image data collected by the camera with each other through parameter conversion. The method is as follows:

[0010] (1) Perform camera internal parameter calibration;

[0011] (2) Perform parameter settings before joint calibration of the camera, including adjusting the position of the lidar, confirming that the published ROS node topic contains the necessary information, and setting the checkerboard parameters;

[0012] (3) Dataset capture: Collect 2D images and 3D point cloud data from the lidar and camera at the same time and position;

[0013] (4) Segment the calibration plate point cloud by coordinate adjustment, collect samples, perform initial estimation and nonlinear optimization to improve accuracy, and complete joint calibration;

[0014] S3. By basically limiting the original values ​​of the contact network geometric parameters measured according to the railway contact network design standards and performing data screening, effective measurement data can be obtained;

[0015] S4. Based on the deep learning model and LiDAR, the positioning and ranging of the catenary itself, the catenary suspension string and the catenary locator are completed; the autonomous obstacle avoidance of the mechanical deicing device is realized;

[0016] S5. By considering the workspace and obstacles, the robot safety range is limited and the robot terminal deicing tool safety range is limited;

[0017] S6. Adjust the position and posture of the robot end tool to remove ice.

[0018] Furthermore, after step (4) in step S2, the step further includes: evaluating the joint calibration result and data fusion reliability by comparing the calibration parameters and the reprojection error index: the specific method is:

[0019] Based on the calibrated camera intrinsic parameters and camera radar extrinsic parameters, the lidar point cloud is projected into a two-dimensional image. If the two-dimensional point of the lidar point cloud projection is within the range of the two-dimensional image, the position of the lidar point is recorded, and the color of the corresponding pixel on the camera image is matched and displayed. Finally, the correctness of the camera radar extrinsic parameter calibration is evaluated; by comparing the calibration parameters and the reprojection error indicators, the alignment of the point cloud and the two-dimensional image is observed, and the average reprojection error less than the predetermined value is taken as the best calibration parameter set result.

[0020] Furthermore, the specific method for limiting the safety range of the robot in step S5 is: before using the robot to operate, the safety speed and effective range of the robot's safe operation are first limited, and the X coordinate of the robot is limited to [X min ,X max ]The fixed value in the middle, X min is the minimum value of the robot end deicing tool moving in the X-axis direction, X max The maximum value of the robot end deicing tool in the X-axis direction; the Y coordinate limit range is the maximum left and right pull-out value of the contact network, which should not be less than the allowable working range Y of the existing contact network. l and maximum working range Y max .

[0021] Furthermore, the specific method for limiting the safety range of the robot in step S5 also includes: the maximum value of the Z coordinate limit range is limited to the maximum value Z of the contact line height. max , the minimum value Z of the Z coordinate range min It should be greater than the vertical height of the three connecting rods after the robot's three-axis shoulder joint returns to zero position and is extended, that is, the safety range of the Z coordinate is [Z min ,Z max ], so far, the range in which the robotic de-icing tool can safely move is a limited three-dimensional space.

[0022] Furthermore, the method for limiting the safety range of the robot terminal deicing tool in S5 is:

[0023] Relative to the forward direction, when it is at the lower left side of the contact line, it rotates counterclockwise, so that the striking piece acts on the lower right side of the contact line;

[0024] Relative to the forward direction, when it is below the right side of the contact line, it rotates clockwise, so that the striking piece acts on the lower left side of the contact line.

[0025] Furthermore, the basic limitation of the original values ​​of the contact network geometric parameters measured in step S3 includes limiting the contact network's conductance value, and the specific method is:

[0026] Basic restrictions on the contact network geometric parameters and contact network height: H min≤H≤H max , H min is the minimum value of the overhead contact network height direction (based on the rail surface), H max It is the maximum value in the overhead contact network height direction (based on the rail surface);

[0027] For the overhead contact geometry parameters:

[0028] H=H 0 +h=H 0 +xcosθ

[0029] z=xsinθ

[0030] Where H is the contact line height, h is the distance from the contact line to the horizontal plane where the non-contact measurement system is located, z is the pull-out value, and H 0 is the height of the laser radar scanning point from the rail surface, θ is the angle between the laser radar center axis and the scanning laser beam;

[0031] The scanning distance x must meet the following requirements before the calculated height conductance value can be used as valid data for subsequent use:

[0032]

[0033] Where θ l It is half of the effective measurement angle of the LiDAR;

[0034] Furthermore, the basic limitation of the original values ​​of the contact network geometric parameter measurement in step S3 also includes limiting the pull-out value of the contact network, and the specific method is:

[0035] The trend of the contact network geometric parameter pullout value is expressed by the formula y=±ax+b. The pullout value has the characteristics that the point at the maximum position is continuous but not differentiable, and the pullout values ​​in other cases are continuous and differentiable.

[0036] Therefore, using the pull-out value and span value at each position on the contact network, the first-order inverse of the pull-out value curve y′=±a is obtained, that is, the pull-out value of the positioning point corresponding to the railway B pillar is bb', the span ac=ab'+cb', then the slope value of the contact line ab is a=bb' / ab', and the slope value of the contact line bc is a=bb' / cb';

[0037] The change in the pull-out value within one measurement cycle is:

[0038]

[0039] Where Δz is the change in the pull-out value within a measurement cycle; Δt is the time change of the cycle; v is the train speed; ac represents the span of a measurement cycle; bb′ represents the maximum value of the pull-out value; z represents the pull-out value;

[0040] The pull-out value Z obtained from the nth measurement n The same pull-out value Z as the last measurement n-1 Must satisfy the formula |Z n -Z n-1 |≤Δz, the pulled value data is considered valid.

[0041] Furthermore, the specific method of positioning and ranging the contact network itself, the contact network dropper and the contact network locator in S4 is as follows:

[0042] Start the main detection program and set the working parameters of the monitoring camera;

[0043] Load the trained image recognition model, the camera automatically focuses the lens and obtains the image and transmits it to the image processing model to obtain the real-time position of the contact network itself, the contact network dropper and the contact network locator;

[0044] Combined with the visual lidar fusion function module, the real-time position is obtained, and the rotation matrix R and translation matrix T from the camera coordinate system to the robot base coordinate system are calculated based on hand-eye calibration. The corresponding positions of the contact network itself, the contact network suspension string and the contact network locator in the robot base system are obtained, and the specific distance is calculated.

[0045] According to a second aspect of the present invention, there is provided a railway overhead line mechanical deicing system, comprising:

[0046] Work preparation unit, used for non-contact measurement system preparation;

[0047] The parameter interconnection unit is used to perform a calibration method based on plane constraints, so that the point cloud data collected by the lidar and the two-dimensional image data collected by the camera are related to each other through parameter conversion. The method is as follows:

[0048] (1) Perform camera internal parameter calibration;

[0049] (2) Perform parameter settings before joint calibration of the camera, including adjusting the position of the lidar, confirming that the published ROS node topic contains the necessary information, and setting the checkerboard parameters;

[0050] (3) Dataset capture: Collect 2D images and 3D point cloud data from the lidar and camera at the same time and position;

[0051] (4) The point cloud of the calibration plate is segmented by coordinate adjustment, samples are collected, and initial estimation and nonlinear optimization are performed to improve the accuracy, thus completing the joint calibration.

[0052] A data screening unit is used to obtain valid measurement data by performing basic restrictions on the original values ​​of the contact network geometric parameters measured according to the railway contact network design standards and performing data screening;

[0053] The obstacle ranging unit is used to complete the positioning and ranging of the catenary itself, the catenary suspension string and the catenary locator based on the deep learning model and the lidar; and realize the autonomous obstacle avoidance of the mechanical deicing device;

[0054] A range limiting unit, used to limit the safety range of the robot and the safety range of the de-icing tool at the end of the robot by considering the workspace and obstacles;

[0055] The robot posture adjustment unit is used for adjusting the position and posture of the robot's end tool for deicing.

[0056] As a third aspect of the present invention, it also relates to a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement a method for mechanical de-icing of a railway contact network.

[0057] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0058] 1. The railway contact network mechanical deicing method of the present invention adopts deicing tools and combines them with a robot platform to build a complete contact network mechanical deicing system, which can realize a series of intelligent operations such as autonomous operation, autonomous following of contact network fluctuations, and autonomous obstacle avoidance. Compared with the traditional manual deicing method, it greatly improves the efficiency of deicing operations and significantly shortens the time required for deicing operations, thereby effectively reducing the duration of railway operation interruptions caused by contact network icing and ensuring the efficient and stable operation of rail transit.

[0059] 2. The railway contact network mechanical deicing method of the present invention realizes the coordinated work of the camera and the laser radar by correlating the point cloud data collected by the laser radar and the two-dimensional image data collected by the camera, and realizes non-contact identification and tracking of the target; so that the present invention can accurately identify the contact network wires and surrounding components such as suspension strings and positioners during the deicing operation, and can accurately calculate the relative position relationship between them and the deicing device in real time. Based on this, the deicing robot can automatically avoid these components smoothly and accurately during operation, greatly reducing the risk probability of collision and interference between the deicing device and the surrounding components of the contact network, effectively reducing the damage caused by collision to the contact network components and the deicing device itself, significantly improving the safety and reliability of deicing operations, extending the overall service life of the contact network and the deicing device, and reducing the maintenance cost and frequency of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A flow chart of an implementation method of a preferred embodiment of the present invention;

[0061] Figure 2This is a schematic diagram of non-contact measurement of overhead wire height and pull-out value in a preferred embodiment of the present invention;

[0062] Figure 3 A schematic top view of a railway overhead contact network according to a preferred embodiment of the present invention;

[0063] Figure 4 This is a schematic diagram of the operation of a mechanical deicing device according to a preferred embodiment of the present invention;

[0064] In all the drawings, the same reference numerals represent the same technical features, specifically: 1-hanging string, 2-locator, 3-contact wire, 4-icing, 5-deicing tool, 6-robotic platform, 7-non-contact measurement system, 8-engineering vehicle platform, 9-sleeper, 10-contact network, 11-rail, 12-positioning device, 13-pillar, 14-feeder. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0066] Example 1

[0067] Please refer to Figure 1 This embodiment provides a method for mechanical deicing of a railway contact network, comprising the following steps:

[0068] Step 1: System preparation

[0069] 1.1 Open the protective cover of the non-contact measurement system

[0070] The protective cover used in this embodiment is used to prevent external dust or foreign matter from damaging the core components inside the non-contact measurement system, and also to prevent the laser emitted by the internal laser radar from harming the human eye. The outside is equipped with a pneumatic push rod cover that can be raised and lowered autonomously. When the non-contact measurement system is used, the cover is pneumatically pushed up. After use, the cover is automatically lowered to protect the device. Rubber gaskets are installed on the four sides of the protective cover to play a sealing and buffering role. The pneumatic actuator of the present invention can also be electric.

[0071] 1.2 Turn on the self-cleaning function of the non-contact measurement system and clean it back and forth twice

[0072] During the initial use of the non-contact measurement system, in order to avoid inaccurate data caused by foreign objects blocking the field of view of the camera or radar, the system can be observed through the remote driver's cab in conjunction with the main camera, and the cleaning tool can be manually or automatically controlled to perform two round trip self-cleaning. The actuators of the vacuum cleaning system can be in a variety of ways, such as air pipe blowing or slide rail brushes. The vacuum cleaning system can be operated by a simple button in the cab, and can effectively clean optical components even during measurement.

[0073] 1.3 Turn on the non-contact measurement system constant temperature and humidity system and preheat for 5 minutes

[0074] Due to the complex external environment of electrified railways, such as extreme cold weather or freezing rain, the non-contact measurement system is equipped with a constant temperature and humidity system. The internal heating and constant humidity system is evenly arranged using resistance wires, that is, a constant temperature adjustment system for the working environment of the detection device, which improves the reliability and service life of the product.

[0075] 1.4 Setting de-icing process parameters

[0076] According to the ice coverage of the contact line on site, such as the shape and thickness of the ice, set the de-icing process parameters, including the speed of the de-icing tool, the moving speed of the robot or three-dimensional motion platform, and the obstacle avoidance safety distance and time.

[0077] Step 2: Joint calibration of mechanical de-icing system

[0078] Different sensors have independent coordinate systems and different acquisition frequencies. During the de-icing operation of the contact network, the de-icing tool at the end of the robot needs to automatically follow the ups and downs of the contact network, and also needs to autonomously avoid components near the contact line. This requires the non-contact measurement system to not only identify the contact line and obstacles (contact network hanging strings and contact network locators), but also to calculate the distance from the identified contact line and obstacles (contact network hanging strings and contact network locators) to the robot. Therefore, the present invention uses a calibration method based on plane constraints to jointly calibrate the laser radar 1 and camera 1 in the non-contact measurement system, and obtain the conversion relationship between the laser radar 1 and camera 1 data, that is, to find the corresponding pixel points in the laser radar point cloud data and the camera two-dimensional image at the same time. The camera radar joint calibration process is as follows:

[0079] 2.1 Camera internal calibration

[0080] The Zhang Zhengyou calibration method is used to calibrate the intrinsic parameters of the camera. Camera 1 is used to take 20 standard chessboard grid images at different angles. The position of each photo relative to camera 1 is different. The photo pixels are 1920x1080. The intrinsic parameter matrix, distortion coefficient and other parameters of the camera are calculated by MATLAB calibration toolbox. Then, standard horizontal markers and lateral position markers are set in front of the camera. The horizontal marker is a level ruler, and the lateral position marker is an obstacle placed 2m above the rail engineering vehicle. Camera 1 is used for image measurement, and the image measurement results are used as feedback information to adjust the horizontal angle and yaw angle of the camera to zero.

[0081] 2.2 Parameter settings before camera joint calibration

[0082] Adjust the installation position of LiDAR 1 so that the scanning point falls in the same field of view as the camera, run the ROS node drivers of LiDAR 1 and Camera 1, and confirm that the published ROS node topic contains the point cloud of LiDAR 1 lidar_topic, the image of Camera 1 camera_topic, and the information of Camera 1 camera_info. The information topic content of Camera 1 camrea_info includes the message header header, matrix D, K, R, P, where D and K are the intrinsic parameters of Camera 1, D is the distortion coefficient of Camera 1, K is the intrinsic parameter matrix of Camera 1, D, K, P are the results of the intrinsic calibration of Camera 1 in step 2.1, R is the rotation matrix, which is set as the unit matrix, that is, P is the corrected projection matrix. Finally, set the chessboard parameters, such as the number of inner corner points, square size, backplane size, and center point offset parameters.

[0083] 2.3 Dataset Extraction

[0084] After starting the camera radar joint calibration main program under the ROS system, open the rviz tool, adjust the distance and posture of the calibration board from the lidar 1 and camera 1, such as a distance range of 1-2m and a calibration board placed at an angle of about 45 degrees to the ground. At the same time, increase the yaw angle and pitch angle of the calibration board to allow the wiring harness of the lidar 1 to hit the calibration board as much as possible. Keep the calibration board still after each change, and then control the main program to collect the 2D images and 3D point clouds of the lidar 1 and camera 1 at the same time and position, ensuring that each recording is more than 6s and contains the topic data of the lidar 1 point cloud lidar_topic, the camera 1 image camera_topic and the camera 1 information camera_info, and a total of about 10 recordings.

[0085] 2.4 Segmentation of calibration plate point cloud and joint calibration

[0086] Use the topic loading tool rosbag of the ROS system to load the previously recorded dataset topic again, and use the rqt_config tool to adjust the xyz coordinate value so that the point cloud on the calibration board is separated from the surrounding environment so that only the chessboard is displayed. If the chessboard is not completely isolated, it may affect the plane fitting of the chessboard and cause a large error in the size of the chessboard. Finally, use the capture tool to repeatedly capture point cloud poses of multiple calibration boards. If the effect of a certain capture is not good, the current capture can be discarded. After collecting more than 20 samples, Xianhu first uses the PnP algorithm and the LM algorithm for initial estimation and nonlinear optimization to improve accuracy, and then solves the transformation relationship between the camera 1 coordinate system and the lidar 1 coordinate system based on the same timestamp. The calculation results include the collected image image, point cloud pcd, pose pose, and the calibrated camera radar external parameter information (i.e., rotation matrix R and translation matrix T).

[0087] 2.5 Evaluation of calibration results and data fusion

[0088] According to the camera intrinsic parameters and camera radar extrinsic parameters calibrated in step 2.4, the lidar point cloud is projected into a two-dimensional image with a resolution of 1920x1080. That is, if the two-dimensional point projected by the lidar point cloud is within the image range, the position of the lidar point is recorded, and the color of the corresponding pixel on the camera image is matched and displayed. Finally, the calibration of the camera radar extrinsic parameters is evaluated to see whether it is correct. This evaluation method compares the calibration parameters and the reprojection error indicators, observes the alignment of the point cloud and the two-dimensional image, and takes the average reprojection error less than 10 mm (5 pixels) as the best calibration parameter set result, thereby ensuring the reliability of the subsequent use of the fused data.

[0089] Step 3: Effective measurement of contact network height and pull-out value

[0090] Please refer to Figure 2-Figure 3After the laser radar 1 and the camera 1 and other sensors are calibrated together in step 2, the non-contact contact measurement system of the contact network composed of the two sensors can be used to measure the geometric parameters of the contact line, such as the height and the pull-out value, and the identification and positioning of the surrounding components, such as the identification and position measurement of the contact network locator and the contact network suspension string, that is, the camera 1 is used for visual recognition to determine whether its components belong to the contact line, the contact network locator or the contact network suspension string. At the same time, the laser radar 1 in the fusion data identifies the target object, such as the contact line, the contact network locator or the contact network suspension string according to the camera head 1, and then accurately measures the target object, such as the contact line, the contact network locator or the contact network suspension string, from the space. The distance to the measurement center point of the contact network non-contact measurement system, these fused measurement data are transmitted to the remote deicing operation main control box by wire or wireless, and then the robot is controlled to perform autonomous deicing operations and autonomously follow the contact line and autonomously avoid obstacles. In order to reduce the amount of calculation data and improve the accuracy of the contact network geometric parameter measurement, the present invention basically limits the original value of the contact network geometric parameter measurement by the railway contact network design standard.

[0091] The first is the contact network geometric parameter height value, the measured value H is min -H max Only valid between min is the minimum value of the overhead contact network height direction (based on the rail surface), H max is the maximum value of the overhead contact network height direction (based on the rail surface); otherwise, the data preprocessing program discards this value and executes the next scanning data cycle. Take the vertical installation method of the overhead contact network non-contact measurement system from the installation plane to the railway rail surface as an example. Figure 2 As shown, the contact line distance measured by laser radar 1 is x and the angle is θ, then

[0092] H=H 0 +h=H 0 +xcosθ

[0093] z=xsinθ

[0094] Where H is the contact line height, h is the distance from the contact line to the horizontal plane where the non-contact measurement system is located, z is the pull-out value, H0 is the height of the laser radar scanning point from the rail surface, and θ is the angle between the central axis of the laser radar and the scanning laser beam.

[0095] Based on the effective range of contact line height conductance Hmin-Hmax, the scanning distance x needs to meet the following requirements in order to use the calculated conductance value as valid data for subsequent use:

[0096]

[0097] Where θ lIt is half of the effective measurement angle of the LiDAR.

[0098] The second is the contact network geometric parameter pull-out value. Since the contact network changes in a zigzag shape, such as Figure 3 As shown, the trend of the pull-out value can be expressed by the formula y=±ax+b. The pull-out value has the characteristic that the point at the maximum position is continuous but not differentiable, and the pull-out values ​​in other cases are continuous and differentiable.

[0099] Therefore, the first-order inverse of the pull-out value curve y′=±a can be obtained by using the pull-out value and span value at each position on the contact network, such as Figure 4 As shown, the pull-out value of the positioning point corresponding to the railway B pillar is bb', the span ac = ab' + cb', then the slope value of the contact line ab is a = bb' / ab', and the slope value of the contact line bc is a = bb' / cb'.

[0100] In addition, the pull-out value can also be limited by whether the periodic change is consistent with the empirical value. Since the contact line changes linearly, the measured value will not undergo a large mutation. Therefore, the accuracy of the pull-out value can be limited by estimating the change in the contact line pull-out value within a measurement cycle. Figure 4 For example, the change in the pull-out value within a measurement cycle is

[0101]

[0102] The pull-out value Z obtained from the nth measurement n The same pull-out value Z as the last measurement n-1 Must satisfy the formula |Z n -Z n-1 |≤Δz, the pull-out value data can be considered valid. For example, if the absolute value of the difference between the pull-out value measured this time and the last measured value exceeds 20mm, the measurement data is considered meaningless.

[0103] Step 4: Autonomous obstacle avoidance of mechanical deicing device

[0104] Since the railway contact network will fluctuate up and down and left and right, a robot with a terminal deicing tool is required to automatically change its position and posture in real time to follow the changes of the contact network during deicing operations. At the same time, it is also necessary to ensure that the remaining components on the contact network, such as the contact network suspension string and wire clamp, the contact network locator and wire clamp, are not damaged during the deicing operation. The mechanical deicing device of the present invention can automatically avoid obstacles and manually avoid obstacles. The manual method is that the driver's cab manually remotely controls the robot away from obstacles through the camera 1 image. The automatic method is that the robot autonomously controls the robot to perform autonomous deicing operations and autonomous obstacle avoidance actions through the environmental data fused by multiple sensors such as laser radar 1 and camera 1 in step 2. According to the driving direction of the engineering vehicle, the main program of the main control box automatically collects the images corresponding to the contact network locator and the contact network suspension string above the engineering vehicle through camera 1 (forward direction) or camera 2 (backward direction), processes the obtained images, and sends them to the deep learning model based on YoloV5 technology to obtain the positions of the contact network suspension string and the contact network locator in the two-dimensional image and mark them out.

[0105] The specific methods for identifying and detecting overhead line obstacles are as follows:

[0106] The first step is to start the main program of the overhead line obstacle visual detection, and set the frame rate, resolution, exposure time, and synchronous fill light time of the detection camera;

[0107] The second step is to turn on the visual inspection function. Enter the main interface of the contact network non-contact inspection system, start the contact network adaptive following function, load the trained recognition model, and the No. 1 autofocus camera acquires the image and transmits it to the image processing device. The model analyzes the real-time position of the contact network itself, the contact network suspension string and the contact network locator, and sets the appropriate fill light brightness during this process. During the contact network inspection process, the host computer displays the inspection video and corresponding recognition results in real time, and records the operation data;

[0108] The third step is the coordinate system transformation after data fusion. According to the aforementioned visual lidar fusion function module, the positions of the contact network itself, the contact network suspension string and the contact network locator in the camera reference system identified by the YoloV5 deep learning model in the second step are calculated using the hand-eye calibration based on nine-point calibration to obtain the rotation matrix R and translation matrix T from the camera coordinate system to the robot base coordinate system, and obtain the corresponding positions of the contact network itself, the contact network suspension string and the contact network locator in the robot base reference system;

[0109] The fourth step is autonomous obstacle avoidance. When the mechanical deicing device encounters the contact network suspension string and contact network locator during the deicing operation, the non-contact detection system identifies the obstacle, and the main program controls the position and posture of the robot's terminal deicing tool to move away from the obstacle. In order to improve the deicing efficiency, the robot's terminal deicing tool can increase the operating speed and effective range of the deicing mechanism at the contact line without obstacles, such as the end deicing tool's swing stick can be on the right side of the contact line. Reduce the operating speed and effective range of the deicing mechanism in places with obstacles, such as the end deicing tool's swing stick is completely on the side and below the contact line. The posture of the robot's terminal deicing tool can also change autonomously with the slope of the contact network to ensure that the direction of the force of the terminal deicing tool is perpendicular to the right side of the contact line, and to ensure that the terminal deicing device does not damage other components while effectively removing ice from the contact network.

[0110] Step 5: Safety range limitation of robot and terminal de-icing tool

[0111] Before using the robot for operation, in order to ensure that the laser radar can effectively scan the zigzag-shaped contact network at the same time, it is necessary to first limit the safe speed and effective range of the robot's safe operation. For example, the robot's X coordinate is limited to [X min , X max ] a fixed value in the middle, X min is the minimum value of the robot end deicing tool moving in the X-axis direction, X max The maximum value of the robot's end deicing tool in the X-axis direction, such as 500mm; the Y coordinate limit range is the maximum left and right pull-out value of the contact network, which should not be less than the allowable working range Y of the existing contact network. 1 and maximum working range Y max , such as Y 1 =950mm, Y max =1250mm, the maximum value of the Z coordinate range is limited to the maximum value of the contact network conductor height Zmax, for example Z max =2500mm. In order to avoid interference between the robot's 3-axis shoulder joint and the contact line, the minimum value of the Z coordinate range is Z min It should be greater than the vertical height of the three connecting rods after the robot's three-axis shoulder joint returns to zero position and is extended, that is, the safety range of the Z coordinate is [Z min , Z max ], such as Z min =1000mm, Z max =2500mm. So far, the range in which the robot de-icing tool can safely move is a limited three-dimensional space.

[0112] After that, it is necessary to limit the safe range of the robot's terminal deicing tool relative to the contact line. Since there are obstacles such as suspension strings and positioners above the contact line, in order to avoid dangerous collisions between the robot and the terminal deicing tool and them, the operating range of the robot's terminal working tool can be further set. The posture of the terminal deicing tool is set to the lower side of the contact line. If it is set to the lower left side (relative to the forward direction), the deicing tool rotates counterclockwise and the knocking piece hits the lower right side of the contact line; if it is set to the lower right side (relative to the forward direction), the deicing tool rotates clockwise and the knocking piece hits the lower left side of the contact line.

[0113] Step 6: Adjust the robot end tool position and posture to remove ice

[0114] The mechanical deicing device of the present invention uses camera images to identify the contact line, contact network locator and contact network suspension string, and obtains the contact network lead height and pull-out value, contact network locator and contact network suspension string real-time position through laser radar, and then controls the robot and the terminal deicing tool to dynamically adjust the position and posture, autonomously track the contact network changes, and when encountering obstacles, such as contact network locator and suspension string, send their position coordinates to the main control system, and then control the robot to stay away from the obstacles, ensuring that no danger occurs during the deicing operation. In addition, the system can also remotely manually and automatically determine the deicing effect through camera 1, and then choose to remotely manually set the deicing knocking speed, or autonomously compare the deicing result before and after the multi-frame image of camera 1, and the main control system autonomously updates the rotation knocking speed.

[0115] Example 2

[0116] This embodiment provides a railway contact network mechanical deicing system, including:

[0117] Work preparation unit, used for non-contact measurement system preparation;

[0118] The parameter interconnection unit is used to perform a calibration method based on plane constraints, so that the point cloud data collected by the lidar and the two-dimensional image data collected by the camera are related to each other through parameter conversion. The method is as follows:

[0119] (1) Perform camera internal parameter calibration;

[0120] (2) Perform parameter settings before joint calibration of the camera, including adjusting the position of the lidar, confirming that the published ROS node topic contains the necessary information, and setting the checkerboard parameters;

[0121] (3) Dataset capture: Collect 2D images and 3D point cloud data from the lidar and camera at the same time and position;

[0122] (4) Segment the calibration plate point cloud by coordinate adjustment, collect samples, perform initial estimation and nonlinear optimization to improve accuracy, and complete joint calibration;

[0123] A data screening unit is used to obtain valid measurement data by performing basic restrictions on the original values ​​of the contact network geometric parameters measured according to the railway contact network design standards and performing data screening;

[0124] The obstacle ranging unit is used to complete the positioning and ranging of the catenary itself, the catenary suspension string and the catenary locator based on the deep learning model and the lidar; and realize the autonomous obstacle avoidance of the mechanical deicing device;

[0125] A range limiting unit, used to limit the safety range of the robot and the safety range of the de-icing tool at the end of the robot by considering the workspace and obstacles;

[0126] The robot posture adjustment unit is used for adjusting the position and posture of the robot's end tool for deicing.

[0127] Example 3

[0128] Please refer to Figure 4 This embodiment provides a railway contact network mechanical deicing device, which includes three parts:

[0129] Component 1 is a railway engineering vehicle platform 8, which provides for the mounting and use of a complete set of mechanical deicing equipment;

[0130] Component 2 is a robot platform 6, which is used to provide the de-icing tool 5 with loading and spatial position and posture adjustment, target object following, autonomous obstacle avoidance and other actions. In order to improve the reliability of the platform, especially for use in low-temperature environments, the mechanism also includes an auxiliary heating system, such as a robot thermal resistance wire protective clothing.

[0131] Component 3 is a deicing tool 5, including a wheel, a knocking actuator (such an actuator may be a rotating motor, a pneumatic motor or an electric spindle, etc.), an end insulating whip, etc. The deicing method is to drive the end whip to inertially knock the ice on the contact line to achieve the deicing effect by rotating the knocking actuator. The rotation direction can be selected automatically according to the forward or backward direction of the engineering vehicle, and the rotating knocking speed can be automatically rotated according to the deicing effect. The end insulating whip can be a multi-section combination.

[0132] Component 4 is a non-contact contact measurement system 7 for the contact network. The main function of this system is to load cameras and laser radars for identifying and measuring contact lines and contact network peripheral components and targets, and tracking and locating them. They are arranged in front and behind the engineering vehicle. For example, the forward system includes laser radar 1 and camera 1, and the backward system includes laser radar 2 and camera 2. Their positions need to be separated from the positions of components 2 and 3 by a certain distance. This distance is related to the speed of the engineering vehicle and the reaction speed of the non-contact measurement system, such as 3m. Depending on the direction of travel of the engineering vehicle, the main control system will automatically choose to use the forward contact network non-contact measurement system or the rear contact network non-contact measurement system. In addition, in order to improve the reliability of the system, it is also equipped with a self-cleaning system, a cold and warm air system, and a safety protection system. The main functions of these three systems are described in the following steps.

[0133] Embodiment 4:

[0134] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned railway contact network mechanical deicing method can be implemented.

[0135] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0136] For an introduction to the computer-readable storage medium provided in this application, please refer to the above method embodiment, and this application will not go into details here.

[0137] It will be easily understood by those skilled in the art that 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 and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for mechanical deicing of railway contact network, characterized in that: include: S1.Prepare the non-contact measurement system; S2. A calibration method based on plane constraints is used to associate the point cloud data collected by the lidar and the two-dimensional image data collected by the camera through parameter conversion. The method is as follows: (1) Perform camera internal parameter calibration; (2) Perform parameter settings before joint calibration of the camera, including adjusting the position of the lidar, confirming that the published ROS node topic contains the necessary information, and setting the checkerboard parameters; (3) Dataset capture: Collect 2D images and 3D point cloud data from the lidar and camera at the same time and position; (4) Segment the calibration plate point cloud by coordinate adjustment, collect samples, perform initial estimation and nonlinear optimization to improve accuracy, and complete joint calibration; S3. By basically limiting the original values ​​of the contact network geometric parameters measured according to the railway contact network design standards and performing data screening, effective measurement data can be obtained; S4. Based on the deep learning model and LiDAR, the positioning and ranging of the catenary itself, the catenary suspension string and the catenary locator are completed; the autonomous obstacle avoidance of the mechanical deicing device is realized; S5. By considering the workspace and obstacles, the robot safety range is limited and the robot terminal deicing tool safety range is limited; S6. Adjust the position and posture of the robot end tool to remove ice.

2. A railway overhead contact network mechanical deicing method according to claim 1, characterized in that: The step S2 further includes after step (4): evaluating the joint calibration result and data fusion reliability by comparing the calibration parameters and the reprojection error index, the specific method being: Based on the calibrated camera intrinsic parameters and camera radar extrinsic parameters, the lidar point cloud is projected into a two-dimensional image. If the two-dimensional point of the lidar point cloud projection is within the range of the two-dimensional image, the position of the lidar point is recorded, and the color of the corresponding pixel on the camera image is matched and displayed. Finally, the correctness of the camera radar extrinsic parameter calibration is evaluated; by comparing the calibration parameters and the reprojection error indicators, the alignment of the point cloud and the two-dimensional image is observed, and the average reprojection error less than the predetermined value is taken as the best calibration parameter set result.

3. A railway overhead contact network mechanical deicing method according to claim 1, characterized in that: The specific method for limiting the safety range of the robot in step S5 is: before using the robot to operate, the safety speed and effective range of the robot's safe operation are limited. The X coordinate of the robot is limited to [X min ,X max ]The fixed value in the middle, X min is the minimum value of the robot end deicing tool moving in the X-axis direction, X max The maximum value of the robot end deicing tool moving in the X-axis direction; the Y coordinate limit range is the maximum left and right pull-out value of the contact network, which should not be less than the allowable working range Y of the existing contact network. l and maximum working range Y max .

4. A railway overhead contact network mechanical deicing method according to claim 3, characterized in that: The specific method for limiting the robot safety range in step S5 also includes: the maximum value of the Z coordinate limit range is limited to the maximum value Z of the contact network guide height max , the minimum value Z of the Z coordinate range min It should be greater than the vertical height of the three connecting rods after the robot's three-axis shoulder joint returns to zero position and is extended, that is, the safety range of the Z coordinate is [Z min ,Z max ], so far, the range in which the robotic de-icing tool can safely move is a limited three-dimensional space.

5. A railway overhead contact network mechanical deicing method according to claim 4, characterized in that: The method for limiting the safety range of the robot end deicing tool in S5 is: Relative to the forward direction, when it is at the lower left side of the contact line, it rotates counterclockwise, so that the striking piece acts on the lower right side of the contact line; Relative to the forward direction, when it is below the right side of the contact line, it rotates clockwise, so that the striking piece acts on the lower left side of the contact line.

6. A railway overhead contact network mechanical deicing method according to claim 1, characterized in that: The basic limitation of the original values ​​of the contact network geometric parameter measurement in step S3 includes limiting the contact network conductance value, and the specific method is: Basic restrictions on the contact network geometric parameters and contact network height: H min ≤H≤H max , H min H is the minimum height of the overhead contact network based on the rail surface. max It is the maximum value of the overhead contact network in the height direction based on the rail surface; For the overhead contact geometry parameters: H=H0+h=H0+xcosθ z=xsinθ Where H is the contact line height, h is the distance from the contact line to the horizontal plane where the non-contact measurement system is located, z is the pull-out value, H0 is the height of the laser radar scanning point from the rail surface, and θ is the angle between the central axis of the laser radar and the scanning laser beam; The scanning distance x needs to meet the following requirements, and the calculated height conductance value is used as valid data for subsequent use: Where θ l It is half of the effective measurement angle of the LiDAR.

7. A railway overhead contact network mechanical deicing method according to claim 5, characterized in that: The basic limitation of the original values ​​of the contact network geometric parameter measurement in step S3 also includes limiting the pull-out value of the contact network, and the specific method is: The trend of the contact network geometric parameter pullout value is expressed by the formula y=±ax+b. The pullout value has the characteristics that the point at the maximum position is continuous but not differentiable, and the pullout values ​​in other cases are continuous and differentiable. Therefore, using the pull-out value and span value at each position on the contact network, the first-order inverse of the pull-out value curve y′=±a is obtained, that is, the pull-out value of the positioning point corresponding to the railway B pillar is bb', the span ac=ab'+cb', then the slope value of the contact line ab is a=bb' / ab', and the slope value of the contact line bc is a=bb' / cb'; The change in the pull-out value within one measurement cycle is: Where Δz is the change in the pull-out value within a measurement cycle; Δt is the time change of the cycle; v is the train speed; ac represents the span of a measurement cycle; bb′ represents the maximum value of the pull-out value; z represents the pull-out value; The pull-out value Z obtained from the nth measurement n The same pull-out value Z as the last measurement n-1 Must satisfy the formula |Z n -Z n-1 |≤Δz, the pulled value data is considered valid.

8. A railway overhead contact network mechanical deicing method according to claim 1, characterized in that: The specific methods for positioning and ranging the overhead contact network itself, the overhead contact network dropper and the overhead contact network locator in S4 are as follows: Start the main detection program and set the working parameters of the monitoring camera; Load the trained image recognition model, the camera automatically focuses the lens and obtains the image and transmits it to the image processing model to obtain the real-time position of the contact network itself, the contact network dropper and the contact network locator; Combined with the visual lidar fusion function module, the real-time position is obtained, and the rotation matrix R and translation matrix T from the camera coordinate system to the robot base coordinate system are calculated based on hand-eye calibration. The corresponding positions of the contact network itself, the contact network suspension string and the contact network locator in the robot base system are obtained, and the specific distance is calculated.

9. A railway contact network mechanical deicing system, characterized in that: include: Work preparation unit, used for non-contact measurement system preparation; The parameter interconnection unit is used to perform a calibration method based on plane constraints, so that the point cloud data collected by the lidar and the two-dimensional image data collected by the camera are related to each other through parameter conversion. The method is as follows: (1) Perform camera internal parameter calibration; (2) Perform parameter settings before joint calibration of the camera, including adjusting the position of the lidar, confirming that the published ROS node topic contains the necessary information, and setting the checkerboard parameters; (3) Dataset capture: Collect 2D images and 3D point cloud data from the lidar and camera at the same time and position; (4) Segment the calibration plate point cloud by coordinate adjustment, collect samples, perform initial estimation and nonlinear optimization to improve accuracy, and complete joint calibration; A data screening unit is used to obtain valid measurement data by performing basic restrictions on the original values ​​of the contact network geometric parameters measured according to the railway contact network design standards and performing data screening; The obstacle ranging unit is used to complete the positioning and ranging of the catenary itself, the catenary suspension string and the catenary locator based on the deep learning model and the lidar; and realize the autonomous obstacle avoidance of the mechanical deicing device; A range limiting unit, used to limit the safety range of the robot and the safety range of the de-icing tool at the end of the robot by considering the workspace and obstacles; The robot posture adjustment unit is used for de-icing by adjusting the position and posture of the robot's end tool.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the method for mechanical deicing of railway contact network as described in claims 1-8.

Images