Material area map updating method and device, storage medium and equipment

By calculating the vehicle's key point position and vehicle body safety frame, identifying the material pile boundary, and updating the global point cloud map after operation, the problem of unmanned loaders mixing the material area map in the station indoors is solved, and bucket collision is avoided and positioning inaccuracy, supporting material margin monitoring.

CN119063715APending Publication Date: 2024-12-03JIANGSU XCMG STATE KEY LAB TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411198462.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing technology has failed to effectively solve the problem of material area map update when unmanned loaders operate indoor mixing stations, especially when identifying material pile point clouds and building operation boundaries, which may cause buckets to collide with the walls and bottom of the silo. , and the material boundary information cannot be updated in time to ensure positioning accuracy.

Method used

By obtaining the global point cloud map before the operation, local point cloud data at the vehicle end and point cloud data in the material area, calculate the vehicle key point position and vehicle body safety box, identify the material pile boundaries and work boundaries, and update the global points after the operation The boundary of cloud map and material pile is used to identify material piles using vertical feature self-attention mechanism and horizontal feature self-attention mechanism, and the boundary points are determined in combination with raster map projection, and the map is updated by screening the change points of material piles.

Benefits of technology

Avoid collision between the bucket and the silo wall during operation, ensure positioning accuracy, and update the point cloud map and material pile boundaries after operation, ensure the matching accuracy of lidar and prior point cloud maps, and support material margin monitor.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119063715A_ABST
    Figure CN119063715A_ABST
Patent Text Reader

Abstract

The invention discloses a material area map updating method and device, a storage medium and equipment. The material area map updating method comprises the steps of performing vehicle multi-point attitude calculation based on a global point cloud map before operation, vehicle end local point cloud data and vehicle attitude data; based on the point cloud data of the material area before operation, carrying out material pile identification and determining a boundary; in response to the fact that the vehicle key point is located in a closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary is not smaller than a preset distance threshold value, operation is completed; updating a global point cloud map according to the global point cloud data after operation, the local point cloud data of the vehicle end and the point cloud data of the material area after operation; and carrying out material pile identification according to the point cloud data of the material area after operation and determining a boundary to obtain a material pile boundary after operation. According to the invention, collision between the bucket and the bin wall caused by material change during operation can be avoided, and the point cloud map and the material pile boundary can be updated after operation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a material area map updating method, device, storage medium and equipment, belonging to the technical field of electronic maps. Background Art

[0002] An unmanned loader is an engineering machinery equipment that is equipped with sensors and controllers on the original loader to achieve unmanned autonomous control. It can be used in closed scenes such as indoor mixing stations to carry out continuous shoveling and transportation operations of sand and stone materials.

[0003] When an unmanned loader is operating in an indoor mixing station, satellite positioning cannot be used because it is a closed indoor scene. A positioning method based on a priori point cloud map can be used. Therefore, before real-time positioning, it is necessary to first build an overall priori point cloud map inside the mixing station.

[0004] Patent CN115269763A discloses a local point cloud map update and maintenance method, device, mobile tool and storage medium. When the latest frame of point cloud data is acquired, the first data information corresponding to each point cloud in the newly acquired frame is determined, and the local point cloud map is updated using the least recently used algorithm based on the first data information.

[0005] Patent CN113776544A discloses a point cloud map update method, device, electronic device and positioning system, which compares the scanned lidar point cloud with the pre-stored point cloud map to determine the difference points between the lidar point cloud and the point cloud map. If the observation probability of the map point is less than a preset value, the point is deleted from the point cloud map to achieve point cloud map update.

[0006] During the shoveling operation of the unmanned loader in the silo, the amount of materials changes continuously, and the material boundary also changes continuously. In order to determine the relationship between the vehicle bucket and the boundary and prevent the bucket from hitting the side walls and the bottom wall of the silo during the shoveling operation, the material boundary information needs to be updated in time. At the same time, when the material changes, the prior map of the current area also needs to be updated to ensure that the lidar scanning results and the current prior point cloud map can be accurately matched to ensure the accuracy of positioning. At the end of the shoveling operation, the material boundary information needs to be updated again and uploaded to the cloud to assist in material balance monitoring.

[0007] The map update method provided by the prior art does not involve the map update of the unmanned loader in the mixing station scene, and does not involve the point cloud recognition of the material pile in the material area, the construction of the operation boundary and the material pile boundary. The prior art is not suitable for updating the point cloud map by screening the material change points. Summary of the invention

[0008] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a material area map update method, device, storage medium and equipment, which can avoid collision between the bucket and the silo wall due to material changes during operation, and can update the point cloud map and the material pile boundary after the operation. To achieve the above purpose, the present invention is implemented by adopting the following technical solutions:

[0009] In a first aspect, the present invention provides a material area map updating method, comprising:

[0010] Obtain the global point cloud map before the operation, the local point cloud data of the vehicle, the vehicle posture data, and the point cloud data of the material area before the operation;

[0011] The multi-point posture of the vehicle is calculated based on the global point cloud map before the operation, the local point cloud data of the vehicle end and the vehicle posture data, and the posture of the key points of the vehicle and the vehicle body safety frame are obtained; the material pile is identified and the boundary is determined based on the point cloud data of the material area before the operation, and the material pile boundary and the operation boundary are obtained;

[0012] In response to the key point of the vehicle being within the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary being not less than a preset distance threshold, the material loading operation is completed;

[0013] Obtain the global point cloud data after the operation, the local point cloud data of the vehicle end, and the point cloud data of the material area after the operation;

[0014] The global point cloud map is updated based on the global point cloud data after the operation, the local point cloud data on the vehicle side, and the point cloud data of the material area after the operation to obtain the global point cloud map after the operation; the material pile is identified and the boundary is determined based on the point cloud data of the material area after the operation to obtain the boundary of the material pile after the operation.

[0015] In combination with the first aspect, optionally, the vehicle multi-point pose calculation is performed based on the global point cloud map before the operation, the local point cloud data on the vehicle side, and the vehicle pose data to obtain the vehicle key point pose and the vehicle body safety frame, including:

[0016] Map matching and positioning are performed based on the global point cloud map before the operation and the local point cloud data on the vehicle side to obtain the global position and posture of the laser radar point where the local point cloud data on the vehicle side is collected;

[0017] The multi-point posture of the vehicle is calculated based on the global posture of the laser radar point and the vehicle posture data to obtain the posture of the key points of the vehicle; wherein the posture of the key points of the vehicle includes the posture of the rear axle center point, the steering point, the front axle center point, the left corner point of the bucket, and the right corner point of the bucket;

[0018] Calculate the vehicle body safety frame according to the key point position of the vehicle; wherein the vehicle body safety frame includes a front vehicle body safety frame and a rear vehicle body safety frame;

[0019] The front vehicle body safety frame is a grid map in the vehicle body coordinate system established with the front axle center point as the origin. Figure 4 The edge coordinates are transformed into the global coordinate system to form a set of coordinate points;

[0020] The rear body safety frame is a grid map in the body coordinate system established with the rear axle center point as the origin. Figure 4 The edge coordinates are converted to the global coordinate system to form a set of coordinate points.

[0021] In combination with the first aspect, optionally, the material pile identification and boundary determination based on the point cloud data of the material area before the operation to obtain the material pile boundary and the operation boundary, the material pile identification includes:

[0022] Perform data downsampling and neighborhood point search on the point cloud data of the material area before the operation to obtain sampling points and neighborhood points;

[0023] Based on the vertical feature self-attention mechanism, the vertical feature relationship between the sampling point and the neighboring point is learned to obtain the vertical feature; based on the horizontal feature self-attention mechanism, the horizontal feature relationship between the sampling point and the neighboring point is learned to obtain the horizontal feature; the vertical feature and the horizontal feature are cascaded to obtain the merged feature vector;

[0024] The merged feature vector is sequentially input into the random pooling layer to strengthen the high probability features and into the upsampling layer for origin mapping to obtain the point cloud features;

[0025] The point cloud features are input into the fully connected layer to convert the relationship between the point cloud features and the labeling results, and the material pile point cloud category label is Points_L, the wall point cloud category label is Points_Q, and the ground point cloud category label is Points_D.

[0026] In combination with the first aspect, optionally, the material pile is identified and the boundary is determined based on the point cloud data of the material area before the operation to obtain the material pile boundary and the operation boundary, and the boundary determination includes:

[0027] Taking the center of the silo as the origin, construct the local grid map Local_Grid_Map;

[0028] Project the stockpile point cloud labeled Points_L and the wall point cloud labeled Points_Q onto the grid map Local_Grid_Map respectively, and accumulate point clouds with different heights but the same plane coordinates in the same grid point;

[0029] Take the grid coordinate P_S of the midpoint of the bottom edge of the grid map as the starting point, and take the end grid points on the left, top, and right of the grid map as the end points P_E in clockwise order to construct multiple straight lines;

[0030] Traverse all grid points on the left, top or right from P_S to P_E, and number them in order from near to far {P_0, P_1, P_2...P_N} according to their distance from the starting point P_S;

[0031] Determine the boundary points from {P_0, P_1, P_2…P_N}, and the judgment conditions are:

[0032] (a) The grid occupancy value of the coordinate point P_N is not 0;

[0033] (b) The coordinate point P_N is within 0.2m, and there are at least 4 points with grid occupancy values ​​not equal to 0, so as to remove isolated false points P_Fake;

[0034] (c) Starting from point P_2, three consecutive points P_N-1, P_N-2, and P_N-3 are blank points with grid occupancy values ​​of 0;

[0035] When the above conditions are met simultaneously, the first-order differential change point of the occupied grid value from P_0 to P_N is the boundary point;

[0036] The material pile point cloud is subjected to raster projection, and the boundary points obtained constitute the material pile boundary; the wall point cloud is subjected to raster projection, and the boundary points obtained constitute the operation boundary.

[0037] In combination with the first aspect, optionally, updating the global point cloud map according to the global point cloud data after the operation, the local point cloud data on the vehicle side, and the point cloud data on the material area after the operation to obtain the global point cloud map after the operation includes:

[0038] According to the global point cloud data after the operation and the local point cloud data on the vehicle side, the global point cloud data structure after the operation is constructed;

[0039] According to the point cloud data of the material area after the operation, the change points of the material pile are screened by projection of the grid maps of three planes; wherein the grid maps of the three planes include XOY grid map, XOZ grid map and YOZ grid map;

[0040] The obtained stockpile change points are updated to the global point cloud data structure after the operation to obtain a global point cloud map after the operation.

[0041] In combination with the first aspect, optionally, screening the stockpile change points through grid map projections of three planes based on the point cloud data of the material area after the operation includes:

[0042] Project point P_n in the point cloud data of the material area after operation vertically onto the XOY grid map, XOZ grid map and YOZ grid map to obtain points A, B and C respectively;

[0043] In response to the change in the grid projection values ​​of points A, B, and C before and after the operation, points A, B, and C are the change points on the current grid map, and the point attribute is [change];

[0044] The set of all change points in the XOY grid map is {P_XOY}, and point A [X_A_Map, Y_A_Map] is a point in the change point set {P_XOY}; the grid map coordinates of point A are [X_A_Grid, Y_A_Grid], and along the direction from point A to point P_n, all elevations are arranged in ascending order [Z_0, Z_1…Z_n];

[0045] The set of all change points in the XOZ grid map is {P_XOZ}, and point B [X_B_Map, Y_B_Map] is a point in the change point set {P_XOZ}; the grid map coordinates of point B are [X_B_Grid, Y_B_Grid], and along the direction from point B to point P_n, all elevations are arranged in ascending order [Y_0, Y_1…Y_n];

[0046] The set of all change points in the YOZ grid map is {P_YOZ}, and point C [X_C_Map, Y_C_Map] is a point in the change point set {P_YOZ}; the grid map coordinates of point C are [X_C_Grid, Y_C_Grid], and along the direction from point C to point P_n, all elevations are arranged in ascending order [X_0, X_1…X_n];

[0047] For point B, let X_B_Map = X_A_Map = X1, traverse, for the grid map coordinates of point B obtained by (X1, Zi), [X_B_Grid, Y_B_Grid], search [Y_0, Y_1…Y_n] to see if there is a value equal to Y_A_Map;

[0048] For point C, let Y_C_Map = Y_A_Map = Y1, traverse, for the grid map coordinates of point C obtained by (Y1, Zi), [X_C_Grid, Y_C_Grid], search [X_0, X_1…X_n] to see if there is a value equal to X_A_Map;

[0049] If: [Y_0, Y_1…Y_n] has a value equal to Y_A_Map and [X_0, X_1…X_n] has a value equal to X_A_Map, then the coordinate (X_A_Map, Y_A_Map, Zi) is the stockpile change point, and the point attribute is set to [reserve].

[0050] In combination with the first aspect, optionally, updating the obtained stockpile change points into the global point cloud data structure after the operation to obtain the global point cloud map after the operation includes:

[0051] Insert the points with the attribute [reserve] into the global point cloud data structure ikd-Tree after the job;

[0052] Traverse all points with the attribute [reserve] and search for the nearest point with the attribute [delete] around each [reserve] point;

[0053] Delete all the nearest points with the attribute [delete] that are searched in the global point cloud data structure ikd-Tree after the job, and obtain the global point cloud map after the job.

[0054] In a second aspect, the present invention provides a material area map updating device, comprising:

[0055] Acquisition module: used to acquire the global point cloud map before the operation, the local point cloud data on the vehicle side, the vehicle posture data, and the point cloud data of the material area before the operation; and used to acquire the global point cloud data after the operation, the local point cloud data on the vehicle side, and the point cloud data of the material area after the operation;

[0056] Material pile identification and boundary construction module: used to calculate the multi-point posture of the vehicle based on the global point cloud map before the operation, the local point cloud data of the vehicle end and the vehicle posture data, and obtain the key point posture of the vehicle and the vehicle body safety frame; based on the point cloud data of the material area before the operation, the material pile is identified and the boundary is determined to obtain the material pile boundary and the operation boundary;

[0057] Boundary relationship determination module: used to respond to the material loading operation being completed when the key point of the vehicle is within the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary is not less than a preset distance threshold;

[0058] Update module: used to update the global point cloud map according to the global point cloud data after the operation, the local point cloud data on the vehicle side and the point cloud data of the material area after the operation, so as to obtain the global point cloud map after the operation; identify the material pile and determine the boundary according to the point cloud data of the material area after the operation, so as to obtain the boundary of the material pile after the operation.

[0059] In a third aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the material area map updating method described in the first aspect.

[0060] In a fourth aspect, the present invention provides a computer device, characterized in that it includes:

[0061] Memory, for storing computer programs / instructions;

[0062] A processor is used to execute the computer program / instructions to implement the steps of the material area map updating method described in the first aspect.

[0063] Compared with the prior art, the material area map updating method, device, storage medium and device provided in the embodiments of the present invention have the following beneficial effects:

[0064] The present invention calculates the multi-point posture of the vehicle based on the global point cloud map before the operation, the local point cloud data on the vehicle side and the vehicle posture data, and obtains the posture of the key points of the vehicle and the vehicle body safety frame; the material pile is identified and the boundary is determined based on the point cloud data of the material area before the operation, and the material pile boundary and the operation boundary are obtained; in response to the key point of the vehicle being in the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary is not less than the preset distance threshold, the material loading operation is completed; during the operation, the present invention first matches the global point cloud map before the operation and the local point cloud data on the vehicle side to ensure the accuracy of positioning; during the operation, the present invention can clarify the operation boundary to avoid collision between the bucket and the silo wall due to material changes;

[0065] The present invention updates the global point cloud map according to the global point cloud data after the operation, the local point cloud data on the vehicle side and the point cloud data of the material area after the operation, so as to obtain the global point cloud map after the operation; the material pile is identified and the boundary is determined according to the point cloud data of the material area after the operation, so as to obtain the boundary of the material pile after the operation; the present invention can update the point cloud map and the boundary of the material pile after the operation; by timely updating the point cloud map, it is ensured that the laser radar point cloud and the prior point cloud map can continue to be accurately matched, thereby ensuring the accuracy of positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 is a flow chart of a material area map updating method in Embodiment 1 of the present invention;

[0067] Figure 2 It is a schematic diagram of the key point position and body safety frame of a vehicle in a material area map updating method in Example 1 of the present invention;

[0068] Figure 3 This is a flow chart of material pile identification in a material area map updating method in Example 1 of the present invention;

[0069] Figure 4 It is a schematic diagram of point cloud features obtained under a vertical feature self-attention mechanism and a horizontal feature self-attention mechanism in a material area map updating method in Example 1 of the present invention;

[0070] Figure 5 It is a schematic diagram of determining boundaries in a material area map updating method in Example 1 of the present invention;

[0071] Figure 6It is a schematic diagram of screening material pile change points in a material area map updating method in Example 1 of the present invention;

[0072] Figure 7 It is a schematic diagram of a material area operation scene in a material area map updating method in Example 1 of the present invention. DETAILED DESCRIPTION

[0073] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0074] Embodiment 1:

[0075] like Figure 1 As shown, this embodiment provides a material area map updating method, including:

[0076] Obtain the global point cloud map before the operation, the local point cloud data of the vehicle, the vehicle posture data, and the point cloud data of the material area before the operation;

[0077] The multi-point posture of the vehicle is calculated based on the global point cloud map before the operation, the local point cloud data of the vehicle end and the vehicle posture data, and the posture of the key points of the vehicle and the vehicle body safety frame are obtained; the material pile is identified and the boundary is determined based on the point cloud data of the material area before the operation, and the material pile boundary and the operation boundary are obtained;

[0078] In response to the key point of the vehicle being within the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary being not less than a preset distance threshold, the material loading operation is completed;

[0079] Obtain the global point cloud data after the operation, the local point cloud data of the vehicle end, and the point cloud data of the material area after the operation;

[0080] The global point cloud map is updated based on the global point cloud data after the operation, the local point cloud data on the vehicle side, and the point cloud data of the material area after the operation to obtain the global point cloud map after the operation; the material pile is identified and the boundary is determined based on the point cloud data of the material area after the operation to obtain the boundary of the material pile after the operation.

[0081] This embodiment is based on Figure 7 The schematic diagram of the material area operation scene shown in the figure discusses a material area map updating method in combination with vehicle operation. The specific implementation steps are as follows.

[0082] Step 1: Obtain the global point cloud map before the operation, the local point cloud data of the vehicle, the vehicle posture data, and the point cloud data of the material area before the operation.

[0083] Figure 7In the schematic diagram of the material area operation scene shown, the starting point position is the initial position for mapping, and the global point cloud map before the operation is the global point cloud map constructed from the starting point.

[0084] Step 2: Calculate the multi-point position and posture of the vehicle based on the global point cloud map before the operation, the local point cloud data of the vehicle, and the vehicle posture data to obtain the key point position and posture of the vehicle and the vehicle body safety frame.

[0085] Step 2.1: Perform map matching and positioning based on the global point cloud map before the operation and the local point cloud data on the vehicle side to obtain the global pose of the point where the laser radar that collected the local point cloud data on the vehicle side is located.

[0086] like Figure 2 As shown, the laser radar is deployed at the top of the vehicle body P6, the rear body IMU is deployed at the center of the rear axle P1, the steering angle sensor is deployed at the hinge point of the front and rear bodies P2, and the front body IMU is deployed at the center of the front axle P3. Figure 2 As shown, the bucket left corner point P4 and the bucket right corner point P5.

[0087] like Figure 2 As shown, the local point cloud data of the vehicle is matched with the global point cloud map to obtain the global pose [X_Lidar, Y_Lidar, Z_Lidar, Yaw_B] of the point P6 where the laser radar that collects the local point cloud data of the vehicle is located.

[0088] Step 2.2: Calculate the multi-point pose of the vehicle based on the global pose of the laser radar point and the vehicle pose data to obtain the pose of the key points of the vehicle.

[0089] like Figure 2 As shown, the key points of the vehicle include the positions of the rear axle center point P1, the steering point P2 (the point where the front and rear vehicle bodies are hinged), the front axle center point P3, the bucket left corner point P4, and the bucket right corner point P5.

[0090] Calculate the pose of the rear axle center point P1. According to the [Roll_B, Pitch_B] of the rear body IMU, in the body coordinate system (front X, left Y, top Z), the rear axle center point P1 is the origin, and the coordinates of the laser radar point P6 relative to the rear axle center are [LtoR_X, LtoR_Y, LtoR_Z]. According to [X_Lidar, Y_Lidar, Z_Lidar, Roll_B, Pitch_B, Yaw_B] and [LtoR_X, LtoR_Y, LtoR_Z], the coordinates of the rear axle center point P1 in the global coordinate system are calculated to be [X_Rear, Y_Rear, Z_Rear].

[0091] Calculate the pose of the turning point P2. According to the [Roll_B, Pitch_B] of the rear body IMU, in the body coordinate system (front X, left Y, top Z), the point P6 where the laser radar is located is the origin, and the coordinates of the turning point P2 relative to the laser radar are [TtoL_X, TtoL_Y, TtoL_Z]. According to [X_Lidar, Y_Lidar, Z_Lidar, Roll_B, Pitch_B, Yaw_B] and [TtoL_X, TtoL_Y, TtoL_Z], the coordinates of the turning point P2 in the global coordinate system are calculated to be [X_Turn, Y_Turn, Z_Turn].

[0092] Calculate the pose of the front axle center point P3. Assume that the steering angle is Turn_Degree, the direction is clockwise, the steering point P2 is in the middle of the front and rear axle centers, the rear vehicle heading angle is Yaw_B, and the front and rear axle center distance is L. Then the front axle center heading angle Yaw_A = Yaw_B - sin-1(0.5sin(Turn_Degree)), according to [X_ Turn, Y_ Turn, Z_ Turn, Roll_A, Pitch_A, Yaw_A] and L, calculate the coordinates of the front axle center P3 [X_Front, Y_Front, Z_Front].

[0093] Calculate the pose of the left corner point P4 and the right corner point P5 of the bucket. Taking the center of the front axle as the origin, in the vehicle coordinate system, the coordinates of the left corner point of the bucket are [LCornerToF_X, LCornerToF_Y, LCornerToF_Z], combined with [X_ Turn, Y_Turn, Z_ Turn, Roll_A, Pitch_A, Yaw_A] to get [LCorner_X, LCorner_Y, LCorner_Z], and similarly get the coordinates of the right corner point P5 of the bucket in the global coordinate system [RCorner_X, RCorner_Y, RCorner_Z].

[0094] Step 2.3: Calculate the vehicle safety frame based on the key point pose of the vehicle.

[0095] like Figure 2 As shown, the vehicle body safety frame includes a front vehicle body safety frame and a rear vehicle body safety frame.

[0096] Step 2.3.1: Calculate the rear body safety frame.

[0097] The rear body safety frame is a grid map of the body coordinate system with the rear axle center point as the origin. Figure 4 The edge coordinates are converted to the global coordinate system to form a set of coordinate points.

[0098] With point P1 as the origin, construct a grid map Box_B in the rear vehicle body coordinate system (front X-left Y-top Z). The grid map resolution is 0.05, the length is L, and the width is H.

[0099] For the grid map Box_B, find the grid map Figure 4 The coordinates of the grid point where the edge is located , combined with [X_Rear, Y_Rear, Z_Rear, Roll_B, Pitch_B, Yaw_B], the coordinates of the grid points in the rear body coordinate system are transformed to the global coordinate system to obtain the rear body safety frame .

[0100] Step 2.3.2: Calculate the front vehicle safety frame.

[0101] The front body safety frame is a grid map in the body coordinate system with the front axle center as the origin. Figure 4 The edge coordinates are converted to the global coordinate system to form a set of coordinate points.

[0102] With point P3 as the origin, construct a grid map Box_F in the vehicle coordinate system (front X-left Y-top Z). The grid map resolution is 0.05, the length is L1, and the width is H1.

[0103] For the grid map Box_F, find the grid map Figure 4 The coordinates of the grid point where the edge is located , combined with [X_Front, Y_Front, Z_Front, Roll_A, Pitch_A, Yaw_A], transform the coordinates of the grid points in the front vehicle coordinate system to the global coordinate system to obtain the front vehicle safety frame .

[0104] Step 3: Based on the point cloud data of the material area before the operation, the material pile is identified and the boundary is determined to obtain the material pile boundary and the operation boundary.

[0105] Figure 7 As shown in the schematic diagram of the material area operation scene, the vehicle moves forward into the silo along the planned trajectory A1. During the driving process, the material pile is identified and the boundary is determined.

[0106] Step 3.1: Identify the material pile based on the point cloud data of the material area before the operation.

[0107] In the material area, point cloud categories are mainly divided into walls, ground, and materials, among which the wall and ground surfaces are regular surfaces, and the material surfaces are irregular surfaces.

[0108] For the classification of regular / irregular surfaces, an attention mechanism can be introduced on the basis of deep learning. The main considerations include: (1) The surface features of regular and irregular surfaces are different, and the horizontal and vertical features need to be clearly defined. (2) The attention mechanism network is used to score and integrate the different surface features of the current point, and then reasonable feature encoding is performed.

[0109] Stockpile identification process Figure 3 As shown, it includes input point cloud, downsampling layer, upsampling layer, fully connected layer, and output point cloud.

[0110] Step 3.1.1: Perform data downsampling and neighborhood point search on the material area point cloud data before the operation to obtain sampling points and neighborhood points. In this embodiment,

[0111] The downsampling method is FPS and the neighborhood point search method is KNN.

[0112] Step 3.1.2: Based on the vertical feature self-attention mechanism, the vertical feature relationship between the sampling point and the neighborhood point is learned to obtain the vertical feature; based on the horizontal feature self-attention mechanism, the horizontal feature relationship between the sampling point and the neighborhood point is learned to obtain the horizontal feature; the vertical feature and the horizontal feature are cascaded to obtain the merged feature vector.

[0113] Specifically, the vertical feature relationship between the sampling point and the neighborhood point is defined as: the cross product of the local feature descriptors and the relative difference of the local feature descriptors.

[0114] It is expressed by the following formula:

[0115] ,

[0116] in, Represents the feature vector of sampling point i and neighborhood point k, represents the local feature descriptor of sampling point i, Represents the local feature descriptor of the neighborhood point k of sampling point i, represents the cross product of the local feature descriptor of the sampling point and the neighborhood point, Represents the relative difference between the local feature descriptors of the sampling point and the neighborhood points, Represents a feature concatenation operation.

[0117] Specifically, the horizontal characteristic relationship between the sampling point and the neighborhood point is defined as: the difference in normal curvature between the sampling point and the neighborhood point.

[0118] It is expressed by the following formula:

[0119] ,

[0120] in, represents the difference in normal curvature between sampling point i and neighborhood point k, represents the normal curvature of sampling point i, Represents the normal curvature of the neighborhood point k of sampling point i.

[0121] The local feature learning process under the vertical feature self-attention mechanism and the horizontal self-attention mechanism is as follows Figure 4 shown.

[0122] For the vertical feature self-attention mechanism, the input item is the vertical feature relationship between the sampling point and the neighborhood point (vertical vector feature), and the 1×1 convolution kernel is used to calculate the vertical feature vector results A, B, and C respectively; the convolution result B is transposed and multiplied with the convolution result C to obtain the similarity between the features; the similarity between the features is calculated by Softmax normalization to obtain the self-attention score of the vertical feature; the self-attention score and the convolution result A are multiplied and added in turn to obtain the output feature T_C after learning.

[0123] For the horizontal feature self-attention mechanism, the input item is the horizontal feature relationship between the sampling point and the neighborhood point (horizontal vector feature), and the vertical feature vector results E, F, and G are obtained by using a 1×1 convolution kernel; the convolution result F is transposed and multiplied with the convolution result E to obtain the similarity between the features; the similarity between the features is calculated by Softmax normalization to obtain the self-attention score of the vertical feature; the self-attention score and the convolution result G are multiplied and added in turn to obtain the output feature T_S after learning.

[0124] The vertical feature T_C and the horizontal feature T_S are cascaded to obtain a merged feature vector.

[0125] Step 3.1.3: Input the merged feature vector into the random pooling layer to enhance the high probability features and then into the upsampling layer for origin mapping to obtain the point cloud features.

[0126] Step 3.1.4: Input the point cloud features into the fully connected layer to convert the relationship between the point cloud features and the labeling results, and obtain the material pile point cloud category label as Points_L, the wall point cloud category label as Points_Q, and the ground point cloud category label as Points_D.

[0127] Step 3.2: Determine the boundary based on the point cloud data obtained by the stockpile identification to obtain the stockpile boundary and the operation boundary.

[0128] The input items are the wall point cloud labeled Points_Q and the material pile point cloud labeled Points_L obtained in step S3. When the material pile is grid-projected, the surface of the distant material pile will be intermittently projected onto the grid map, which will affect the boundary construction, such as Figure 5As shown, therefore, the working boundary and the stockpile boundary need to be calculated separately.

[0129] Step 3.2.1: Build a local grid map.

[0130] Taking the center of the silo as the origin, construct the local grid map Local_Grid_Map.

[0131] In this embodiment, the grid map resolution is 0.05 m, and the length and width are 1.2 times the size of the silo.

[0132] Step 3.2.2: Construct the job boundary.

[0133] Step 3.2.2.1: Project the wall point cloud labeled Points_Q onto the grid map Local_Grid_Map, and accumulate point clouds with different heights but the same plane coordinates within the same grid point.

[0134] Step 3.2.2.2: Take the grid coordinate P_S of the midpoint of the bottom edge of the grid map as the starting point, and construct multiple straight lines in a clockwise order with the end grid points on the left, top, and right sides of the grid map as the end points P_E.

[0135] Step 3.2.2.3: Traverse all grid points on the left, top or right from P_S to P_E, and number them in order from near to far {P_0, P_1, P_2...P_N} according to their distance from the starting point P_S.

[0136] Step 3.2.2.4: Determine the boundary points from {P_0, P_1, P_2…P_N}, and the judgment conditions are:

[0137] (a) The grid occupancy value of the coordinate point P_N is not 0;

[0138] (b) The coordinate point P_N is within 0.2m, and there are at least 4 points with grid occupancy values ​​not equal to 0, so as to remove isolated false points P_Fake;

[0139] (c) Starting from point P_2, three consecutive points P_N-1, P_N-2, and P_N-3 are blank points with grid occupancy values ​​of 0;

[0140] When the above conditions are met at the same time, the first-order differential change point of the occupied grid value from P_0 to P_N is the boundary point.

[0141] Step 3.2.2.5: Perform raster projection on the wall point cloud, and the obtained boundary points constitute the work boundary {Boundary2, Boundary3, Boundary4}.

[0142] The attributes of each boundary point of the job boundary include: boundary attribute [First], boundary direction [Up / Left / Right], and global coordinates of the grid point [X, Y].

[0143] Step 3.2.3: Construct the stockpile boundary.

[0144] Step 3.2.3.1: Project the stockpile point cloud labeled Points_L onto the grid map Local_Grid_Map, and accumulate point clouds with different heights but the same plane coordinates within the same grid point.

[0145] Step 3.2.3.2: Take the grid coordinate P_S of the midpoint of the bottom edge of the grid map as the starting point, and construct multiple straight lines in a clockwise order with the end grid points on the left, top, and right sides of the grid map as the end points P_E.

[0146] Step 3.2.3.3: Traverse all grid points on the left, top or right from P_S to P_E, and number them in order from near to far {P_0, P_1, P_2...P_N} according to their distance from the starting point P_S.

[0147] Step 3.2.3.4: Determine the boundary points from {P_0, P_1, P_2…P_N}, and the judgment conditions are:

[0148] (a) The grid occupancy value of the coordinate point P_N is not 0;

[0149] (b) The coordinate point P_N is within 0.2m, and there are at least 4 points with grid occupancy values ​​not equal to 0, so as to remove isolated false points P_Fake;

[0150] (c) Starting from point P_2, three consecutive points P_N-1, P_N-2, and P_N-3 are blank points with grid occupancy values ​​of 0;

[0151] When the above conditions are met at the same time, the first-order differential change point of the occupied grid value from P_0 to P_N is the boundary point.

[0152] Step 3.2.3.5: Perform raster projection on the stockpile point cloud, and the obtained boundary points constitute the stockpile boundary {Boundary1}.

[0153] The attributes of each boundary point of the stockpile boundary include: boundary attribute [Second], boundary direction [Down], and global coordinates of the grid point [X, Y].

[0154] Step 4: In response to the key point of the vehicle being within the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary being not less than a preset distance threshold, the material loading operation is completed.

[0155] In order to ensure that the bucket is full of materials, the position relationship between the bucket and the boundary of the pile needs to be determined. If the bucket is outside the boundary of the pile, the unmanned loader needs to continue moving forward to ensure that the bucket is completely inside the pile. At the same time, the position relationship between the bucket and the first boundary needs to be determined to prevent the bucket from colliding with the first boundary.

[0156] In this embodiment, the key points of the vehicle for determination are the front axle center point P3, the bucket left corner point P4, and the bucket right corner point P5, and the closed area is a closed circle formed by the working boundary and the stockpile boundary.

[0157] Specifically, Figure 7 As shown, after all boundaries (stockpile boundary + operation boundary) {Boundary1, Boundary2, Boundary3, Boundary4} are determined, the relationship between the front axle center point P3, the bucket left corner point P4, the bucket right corner point P5 and the boundary is determined; at the same time, it is determined whether the front vehicle body safety frame Box_F, the rear vehicle body safety frame Box_B and the operation boundary {Boundary2, Boundary3, Boundary4} collide.

[0158] The key points of the vehicle (front axle center point P3, bucket left corner point P4, bucket right corner point P5) are within the closed area, and the conditions that need to be met include:

[0159] Condition (a) The front axle center point P3, the bucket left corner point P4, and the bucket right corner point P5 are all located within the closed circle formed by the working boundary and the stockpile boundary.

[0160] Specifically, take the front axis center point P3 as an example. On the grid map Local_Grid_Map, take P3 as the seed growth point and grow in the eight-neighborhood direction. If the growth point satisfies the following conditions at the same time during the growth process:

[0161] Condition 1: The boundary attribute of the grid where the growth point is located is First.

[0162] Condition 2: The boundary direction attributes of the grid where the growth point is located include [Up], [Left], and [Right].

[0163] Condition 3: The boundary attribute of the grid where the growth point is located is Second,

[0164] Condition 4: The boundary direction attribute of the grid where the growth point is located contains [Down];

[0165] Then P3 is located in the closed circle formed by the working boundary and the stockpile boundary. The relationship between P4, P5 and the closed area is determined in the same way.

[0166] Condition (b) Based on the closest point search method, any point on the front vehicle safety box Box_F , any point on the rear body safety frame Box_B The distance to any coordinate point on the working boundary {Boundary2, Boundary3, Boundary4} shall not be less than 0.5m.

[0167] When conditions (a) and (b) are met at the same time, it is determined that the key points of the vehicle (front axle center point P3, bucket left corner point P4, bucket right corner point P5) are within the closed area, shoveling is completed, and the material loading operation is completed.

[0168] like Figure 7 As shown in the figure, when the key points of the vehicle (front axle center point P3, bucket left corner point P4, bucket right corner point P5) are in the closed area, and the distance between the front and rear safety frames and any coordinate point of the operation boundary is not less than 0.5m, the current material loading operation is determined to be completed, and the vehicle follows the A2 track and reverses out of the silo.

[0169] Step 5: Obtain the global point cloud data after the operation, the local point cloud data of the vehicle end, and the point cloud data of the material area after the operation.

[0170] In this embodiment, it is acquired after the vehicle leaves the silo.

[0171] Step 6: Update the global point cloud map based on the global point cloud data after the operation, the local point cloud data on the vehicle side, and the point cloud data on the material area after the operation to obtain the global point cloud map after the operation.

[0172] Step 6.1: Construct the post-operation global point cloud data structure based on the post-operation global point cloud data and the vehicle-side local point cloud data.

[0173] Step 6.1.1: Build the data structure of the global point cloud map based on IKD-Tree, and set all point attributes to [delete].

[0174] Step 6.1.2: Based on the current vehicle position, intercept the local point cloud map Local_Points_Map with length L, width M, and height H. Construct three plane grid maps respectively, and the attributes in each grid map include [point cloud world coordinates, grid coordinates, elevation point array, point attributes (change point change / non-change point normal), grid projection value].

[0175] The three-plane grid maps include XOY grid map, XOZ grid map and YOZ grid map.

[0176] Step 6.1.3: The local point cloud map Local_Points_Map is projected onto the grid maps of the three planes respectively as a 2D prior map.

[0177] Step 6.2: Based on the point cloud data of the material area after the operation, the change points of the material pile are screened through the grid map projection of three planes.

[0178] Specifically, after shoveling is completed, the shape of the material pile changes, and it is necessary to filter out the change points from the currently scanned point cloud and update them to the global point cloud map.

[0179] According to the point cloud data of the material area after the operation, the change points of the material pile are screened through the grid map projection of three planes. Figure 6 As shown, the specific steps include:

[0180] Step 6.2.1: Project point P_n in the point cloud data of the material area after operation vertically onto the XOY grid map, XOZ grid map and YOZ grid map to obtain points A, B and C respectively.

[0181] Step 6.2.2: In response to the change in the grid projection values ​​of points A, B, and C before and after the operation, points A, B, and C are the change points on the current grid map, and the point attribute is [change].

[0182] Step 6.2.3: The set of all change points in the XOY grid map is {P_XOY}, point A [X_A_Map, Y_A_Map] is a point in the change point set {P_XOY}; the grid map coordinates of point A are [X_A_Grid, Y_A_Grid], and along the direction from point A to point P_n, all elevations are arranged in ascending order [Z_0, Z_1…Z_n].

[0183] The set of all change points in the XOZ grid map is {P_XOZ}, point B [X_B_Map, Y_B_Map] is a point in the change point set {P_XOZ}; the grid map coordinates of point B are [X_B_Grid, Y_B_Grid], and along the direction from point B to point P_n, all elevations are arranged in ascending order [Y_0, Y_1…Y_n].

[0184] The set of all change points in the YOZ grid map is {P_YOZ}, point C [X_C_Map, Y_C_Map] is a point in the change point set {P_YOZ}; the grid map coordinates of point C are [X_C_Grid, Y_C_Grid], and along the direction from point C to point P_n, all elevations are arranged in ascending order [X_0, X_1…X_n].

[0185] Step 6.2.4: For point B, let X_B_Map = X_A_Map = X1, traverse, and for the grid map coordinates of point B obtained by (X1, Zi), [X_B_Grid, Y_B_Grid], search [Y_0, Y_1…Y_n] to see if there is a value equal to Y_A_Map.

[0186] For point C, let Y_C_Map = Y_A_Map = Y1, traverse, and for the grid map coordinates of point C obtained by (Y1, Zi), [X_C_Grid, Y_C_Grid], search [X_0, X_1…X_n] to see if there is a value equal to X_A_Map.

[0187] Step 6.2.5: If: [Y_0, Y_1…Y_n] has a value equal to Y_A_Map and [X_0, X_1…X_n] has a value equal to X_A_Map, then the coordinate (X_A_Map, Y_A_Map, Zi) is the stockpile change point, and the point attribute is set to [reserve].

[0188] Step 6.3: Update the obtained stockpile change points to the global point cloud data structure after the operation to obtain the global point cloud map after the operation.

[0189] Step 6.3.1: Insert the points with the attribute [reserve] into the global point cloud data structure ikd-Tree after the job.

[0190] Step 6.3.2: Traverse all points with the attribute [reserve] and search for the nearest point with the attribute [delete] around each [reserve] point.

[0191] Step 6.3.3: Delete all the nearest points with the attribute [delete] that are searched in the global point cloud data structure ikd-Tree after the job, and obtain the global point cloud map after the job.

[0192] Step 7: Identify the material pile and determine the boundary based on the point cloud data of the material area after the operation to obtain the boundary of the material pile after the operation.

[0193] Specifically, since the boundary of the material pile changes after the material loading operation is completed, it is necessary to identify the material pile again and determine the boundary, determine the set of boundary points of the material pile after the operation, and obtain the boundary Points_Boundary_Update{P_i|i=1,2…n} of the material pile after the operation.

[0194] like Figure 7 As shown, after the material loading operation is completed, the vehicle returns to the starting position along the reverse track A2, and then drives to the unloading port 1 along the forward track A3. After the unloading is completed, it returns to the starting position along the reverse track A4, completing a round of operation cycle.

[0195] Furthermore, it also includes uploading the stockpile boundary Points_Boundary_Update{P_i|i=1,2…n} after the operation to the cloud.

[0196] The present embodiment provides a material area map updating method with steps 1-4. During operation, the global point cloud map before the operation and the local point cloud data on the vehicle side are first matched to ensure the accuracy of positioning. When the vehicle is performing shoveling operations, the operation boundaries are clearly defined to avoid collisions between the bucket and the silo wall due to material changes.

[0197] The material area map updating method provided in this embodiment has steps 5-8, which updates the global point cloud map and the material pile boundary after the vehicle operation, and can also be uploaded to the cloud to provide a reference for material balance monitoring; by timely updating the point cloud map, it is ensured that the lidar point cloud and the prior point cloud map can continue to accurately match, thereby ensuring the accuracy of positioning.

[0198] Embodiment 2:

[0199] This embodiment provides a material area map updating device, including:

[0200] Acquisition module: used to acquire the global point cloud map before the operation, the local point cloud data on the vehicle side, the vehicle posture data, and the point cloud data of the material area before the operation; and used to acquire the global point cloud data after the operation, the local point cloud data on the vehicle side, and the point cloud data of the material area after the operation;

[0201] Material pile identification and boundary construction module: used to calculate the multi-point posture of the vehicle based on the global point cloud map before the operation, the local point cloud data of the vehicle end and the vehicle posture data, and obtain the key point posture of the vehicle and the vehicle body safety frame; based on the point cloud data of the material area before the operation, the material pile is identified and the boundary is determined to obtain the material pile boundary and the operation boundary;

[0202] Boundary relationship determination module: used to respond to the material loading operation being completed when the key point of the vehicle is within the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary is not less than a preset distance threshold;

[0203] Update module: used to update the global point cloud map according to the global point cloud data after the operation, the local point cloud data on the vehicle side and the point cloud data of the material area after the operation, so as to obtain the global point cloud map after the operation; identify the material pile and determine the boundary according to the point cloud data of the material area after the operation, so as to obtain the boundary of the material pile after the operation.

[0204] Embodiment 3:

[0205] This embodiment provides a computer-readable storage medium on which a computer program / instruction is stored, characterized in that when the computer program / instruction is executed by a processor, the steps of the material area map updating method described in Example 1 are implemented.

[0206] Embodiment 4:

[0207] This embodiment provides a computer device, characterized by comprising:

[0208] Memory, for storing computer programs / instructions;

[0209] A processor is used to execute the computer program / instructions to implement the steps of the material area map updating method described in Example 1.

[0210] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0211] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0212] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0213] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0214] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A material area map updating method, characterized in that: include: Obtain the global point cloud map before the operation, the local point cloud data of the vehicle, the vehicle posture data, and the point cloud data of the material area before the operation; The multi-point posture of the vehicle is calculated based on the global point cloud map before the operation, the local point cloud data of the vehicle end and the vehicle posture data, and the posture of the key points of the vehicle and the vehicle body safety frame are obtained; the material pile is identified and the boundary is determined based on the point cloud data of the material area before the operation, and the material pile boundary and the operation boundary are obtained; In response to the key point of the vehicle being within the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary being not less than a preset distance threshold, the material loading operation is completed; Obtain the global point cloud data after the operation, the local point cloud data of the vehicle end, and the point cloud data of the material area after the operation; The global point cloud map is updated based on the global point cloud data after the operation, the local point cloud data on the vehicle side, and the point cloud data of the material area after the operation to obtain the global point cloud map after the operation; the material pile is identified and the boundary is determined based on the point cloud data of the material area after the operation to obtain the boundary of the material pile after the operation.

2. The material area map updating method according to claim 1, characterized in that: The multi-point position and posture calculation of the vehicle is performed based on the global point cloud map before the operation, the local point cloud data of the vehicle end and the vehicle posture data to obtain the key point position and posture of the vehicle and the vehicle body safety frame, including: Map matching and positioning are performed based on the global point cloud map before the operation and the local point cloud data on the vehicle side to obtain the global position and posture of the laser radar point where the local point cloud data on the vehicle side is collected; The multi-point posture of the vehicle is calculated based on the global posture of the laser radar point and the vehicle posture data to obtain the posture of the key points of the vehicle; wherein the posture of the key points of the vehicle includes the posture of the rear axle center point, the steering point, the front axle center point, the left corner point of the bucket, and the right corner point of the bucket; Calculate the vehicle body safety frame according to the position and posture of the key points of the vehicle; wherein the vehicle body safety frame includes a front vehicle body safety frame and a rear vehicle body safety frame; The front vehicle body safety frame is a set of coordinate points formed by establishing a grid map in a vehicle body coordinate system with the front axle center point as the origin and converting the four-side coordinates of the grid map into a global coordinate system; The rear vehicle body safety frame is a set of coordinate points formed by establishing a grid map in a vehicle body coordinate system with the rear axle center point as the origin and converting the four-side coordinates of the grid map into a global coordinate system.

3. The material area map updating method according to claim 1, characterized in that: The material pile identification and boundary determination based on the material area point cloud data before the operation to obtain the material pile boundary and the operation boundary, the material pile identification includes: Perform data downsampling and neighborhood point search on the point cloud data of the material area before the operation to obtain sampling points and neighborhood points; Based on the vertical feature self-attention mechanism, the vertical feature relationship between the sampling point and the neighboring point is learned to obtain the vertical feature; based on the horizontal feature self-attention mechanism, the horizontal feature relationship between the sampling point and the neighboring point is learned to obtain the horizontal feature; the vertical feature and the horizontal feature are cascaded to obtain the merged feature vector; The merged feature vector is sequentially input into the random pooling layer to strengthen the high probability features and into the upsampling layer for origin mapping to obtain the point cloud features; The point cloud features are input into the fully connected layer to convert the relationship between the point cloud features and the labeling results, and the material pile point cloud category label is Points_L, the wall point cloud category label is Points_Q, and the ground point cloud category label is Points_D.

4. The material area map updating method according to claim 3, characterized in that: The material pile is identified and the boundary is determined based on the point cloud data of the material area before the operation to obtain the material pile boundary and the operation boundary. The boundary determination includes: Taking the center of the silo as the origin, construct the local grid map Local_Grid_Map; Project the stockpile point cloud labeled Points_L and the wall point cloud labeled Points_Q onto the grid map Local_Grid_Map respectively, and accumulate point clouds with different heights but the same plane coordinates in the same grid point; Take the grid coordinate P_S of the midpoint of the bottom edge of the grid map as the starting point, and take the end grid points on the left, top, and right of the grid map as the end points P_E in clockwise order to construct multiple straight lines; Traverse all grid points on the left, top or right from P_S to P_E, and number them in order from near to far {P_0, P_1, P_2...P_N} according to their distance from the starting point P_S; Determine the boundary points from {P_0, P_1, P_2…P_N}, and the judgment conditions are: (a) The grid occupancy value of the coordinate point P_N is not 0; (b) The coordinate point P_N is within 0.2m, and there are at least 4 points with grid occupancy values ​​not equal to 0, so as to remove isolated false points P_Fake; (c) Starting from point P_2, three consecutive points P_N-1, P_N-2, and P_N-3 are blank points with grid occupancy values ​​of 0; When the above conditions are met simultaneously, the first-order differential change point of the occupied grid value from P_0 to P_N is the boundary point; The material pile point cloud is subjected to raster projection, and the boundary points obtained constitute the material pile boundary; the wall point cloud is subjected to raster projection, and the boundary points obtained constitute the operation boundary.

5. The material area map updating method according to claim 1, characterized in that: The updating of the global point cloud map according to the global point cloud data after the operation, the local point cloud data of the vehicle end and the point cloud data of the material area after the operation to obtain the global point cloud map after the operation includes: According to the global point cloud data after the operation and the local point cloud data on the vehicle side, the global point cloud data structure after the operation is constructed; According to the point cloud data of the material area after the operation, the change points of the material pile are screened by projection of the grid maps of three planes; wherein the grid maps of the three planes include XOY grid map, XOZ grid map and YOZ grid map; The obtained stockpile change points are updated to the global point cloud data structure after the operation to obtain a global point cloud map after the operation.

6. The material area map updating method according to claim 5, characterized in that: The method of screening the stockpile change points by grid map projection of three planes based on the point cloud data of the material area after the operation includes: Project point P_n in the point cloud data of the material area after operation vertically onto the XOY grid map, XOZ grid map and YOZ grid map to obtain points A, B and C respectively; In response to the change in the grid projection values ​​of points A, B, and C before and after the operation, points A, B, and C are the change points on the current grid map, and the point attribute is [change]; The set of all change points in the XOY grid map is {P_XOY}, and point A [X_A_Map, Y_A_Map] is a point in the change point set {P_XOY}; the grid map coordinates of point A are [X_A_Grid, Y_A_Grid], and along the direction from point A to point P_n, all elevations are arranged in ascending order [Z_0, Z_1…Z_n]; The set of all change points in the XOZ grid map is {P_XOZ}, and point B [X_B_Map, Y_B_Map] is a point in the change point set {P_XOZ}; the grid map coordinates of point B are [X_B_Grid, Y_B_Grid], and along the direction from point B to point P_n, all elevations are arranged in ascending order [Y_0, Y_1…Y_n]; The set of all change points in the YOZ grid map is {P_YOZ}, and point C [X_C_Map, Y_C_Map] is a point in the change point set {P_YOZ}; the grid map coordinates of point C are [X_C_Grid, Y_C_Grid], and along the direction from point C to point P_n, all elevations are arranged in ascending order [X_0, X_1…X_n]; For point B, let X_B_Map = X_A_Map = X1, traverse, for the grid map coordinates of point B obtained by (X1, Zi), [X_B_Grid, Y_B_Grid], search [Y_0, Y_1…Y_n] to see if there is a value equal to Y_A_Map; For point C, let Y_C_Map = Y_A_Map = Y1, traverse, for the grid map coordinates of point C obtained by (Y1, Zi), [X_C_Grid, Y_C_Grid], search [X_0, X_1…X_n] to see if there is a value equal to X_A_Map; If: [Y_0, Y_1…Y_n] has a value equal to Y_A_Map and [X_0, X_1…X_n] has a value equal to X_A_Map, then the coordinate (X_A_Map, Y_A_Map, Zi) is the stockpile change point, and the point attribute is set to [reserve].

7. The material area map updating method according to claim 6, characterized in that: The step of updating the obtained stockpile change points into the global point cloud data structure after the operation to obtain the global point cloud map after the operation includes: Insert the points with the attribute [reserve] into the global point cloud data structure ikd-Tree after the job; Traverse all points with the attribute [reserve] and search for the nearest point with the attribute [delete] around each [reserve] point; Delete all the nearest points with the attribute [delete] that are searched in the global point cloud data structure ikd-Tree after the job, and obtain the global point cloud map after the job.

8. A material area map updating device, characterized in that: include: Acquisition module: used to obtain the global point cloud map before operation, local point cloud data of the vehicle, vehicle posture data and point cloud data of the material area before operation; And used to obtain the global point cloud data after the operation, the local point cloud data of the vehicle end, and the point cloud data of the material area after the operation; Material pile identification and boundary construction module: used to calculate the multi-point posture of the vehicle based on the global point cloud map before the operation, the local point cloud data of the vehicle end and the vehicle posture data, and obtain the key point posture of the vehicle and the vehicle body safety frame; based on the point cloud data of the material area before the operation, the material pile is identified and the boundary is determined to obtain the material pile boundary and the operation boundary; Boundary relationship determination module: used to respond to the material loading operation being completed when the key point of the vehicle is within the closed area formed by the material pile boundary and the operation boundary and the distance between any point on the vehicle body safety frame and the operation boundary is not less than a preset distance threshold; Update module: used to update the global point cloud map according to the global point cloud data after the operation, the local point cloud data on the vehicle side and the point cloud data of the material area after the operation, so as to obtain the global point cloud map after the operation; identify the material pile and determine the boundary according to the point cloud data of the material area after the operation, so as to obtain the boundary of the material pile after the operation.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the material area map updating method described in any one of claims 1-7 are implemented.

10. A computer device, characterized in that: include: Memory, for storing computer programs / instructions; A processor, configured to execute the computer program / instructions to implement the steps of the material area map updating method according to any one of claims 1 to 7.