A method for optimizing a perception system for an open-pit mine truck

Abstract

The application provides a perception system optimization method for an open-pit mine vehicle, and relates to the technical field of perception system optimization, and comprises the following steps: based on a perception system arranged on an open-pit mine vehicle, real-time perception capturing is performed on the surrounding information in the driving range of the same open-pit mine vehicle; based on the real-time perception capturing result, a surrounding perception scene of the same open-pit mine vehicle is constructed; the surrounding perception scene is matched with a preset database to determine a scene itself matching object under the capturing boundary of the surrounding perception scene; the scene itself matching object is evaluated, if the evaluation result meets a standard evaluation condition, it is determined that the perception system does not need to be optimized; otherwise, a to-be-optimized item of the perception system is determined and optimization is performed. Through accurate evaluation of the scene matching link of the perception system, the deficiency of the perception system is accurately found, so that the optimized perception system is improved, and the normal work of the open-pit mine vehicle is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of perception system optimization, in particular to a perception system optimization method for open-pit mine trucks. BACKGROUND

[0002] At present, with the development of science and technology, technology can replace many high-risk jobs, reduce labor costs and ensure safety, including open-pit mining. In the process of open-pit mining, the basic construction investment of mine transportation accounts for about 60% of the total investment of mine construction. In transportation, the automatic driving of open-pit trucks is very important. The perception system, as the basis for vehicle path planning, is a crucial link. However, the current perception system still has shortcomings, and the perceived information cannot accurately match the scene, making optimal planning of the driving route and obstacle avoidance route.

[0003] Therefore, the present application provides a perception system optimization method for open-pit mine trucks. SUMMARY

[0004] The present application provides a perception system optimization method for open-pit mine trucks, which captures the surrounding information in the driving range of the same open-pit mine truck in real time through the perception system installed on the open-pit mine truck, constructs the surrounding perception scene of the same open-pit mine truck, matches the scene with a preset database, determines the scene itself matching object under the capture boundary of the surrounding perception scene, and evaluates it. If the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized, otherwise, the optimization items of the perception system are determined and optimized, the scene matching link of the perception system is accurately evaluated, and the shortcomings of the perception system are accurately found, so as to perfect the optimized perception system and ensure the normal work of the open-pit mine truck.

[0005] The present application provides a perception system optimization method for open-pit mine trucks, which includes:

[0006] Step 1: capturing the surrounding information in the driving range of the same open-pit mine truck in real time based on the perception system installed on the open-pit mine truck;

[0007] Step 2: constructing the surrounding perception scene of the same open-pit mine truck based on the real-time perception capture result;

[0008] Step 3: matching the surrounding perception scene with a preset database to determine the scene itself matching object under the capture boundary of the surrounding perception scene;

[0009] Step 4: evaluating the scene itself matching object, if the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized;

[0010] Otherwise, determine the items to be optimized of the perception system and optimize.

[0011] Preferably, the present application provides a perception system optimization method for a strip mine vehicle, based on the perception system arranged on the strip mine vehicle, real-time perception and capture of the surrounding information within the driving range of the same strip mine vehicle, comprising:

[0012] Determine the driving range of the strip mine vehicle according to the type of the strip mine vehicle;

[0013] Real-time positioning of the current position of the strip mine vehicle, when the positioning result overlaps with the driving range, control the perception system on the strip mine vehicle to start working;

[0014] Based on the perception system, real-time perception and capture of the surrounding information of the strip mine vehicle, wherein the real-time perception and capture result is the point cloud data of each space point of the surrounding space of the strip mine vehicle.

[0015] Preferably, the present application provides a perception system optimization method for a strip mine vehicle, based on the real-time perception and capture result, during the process of constructing the surrounding perception scene of the same strip mine vehicle, comprising:

[0016] Layer extraction of the real-time perception and capture result of the same strip mine vehicle according to the same longitudinal coordinate;

[0017] Determine the absolute value of the gray scale difference of adjacent point clouds in each point cloud layer, when the absolute value of the gray scale difference is greater than the preset edge difference, lock the final point cloud according to max(x01, x02), wherein x01 represents the first point cloud in the adjacent point cloud; x02 represents the second point cloud in the adjacent point cloud; max represents the maximum value symbol;

[0018] Based on all the final point clouds of the same point cloud layer, obtain the layer contour;

[0019] Combine all the layer contours in the order of longitudinal coordinate to construct the three-dimensional scene of the same strip mine vehicle.

[0020] Preferably, the present application provides a perception system optimization method for a strip mine vehicle, based on the real-time perception and capture result, constructing the surrounding perception scene of the same strip mine vehicle, comprising:

[0021] Based on the vehicle structure parameters of the same strip mine vehicle, obtain the three-dimensional mine vehicle of the same strip mine vehicle;

[0022] Based on the three-dimensional mine vehicle and the corresponding three-dimensional scene, construct the surrounding perception scene of the same strip mine vehicle.

[0023] Preferably, the all-around perception scene is matched with a preset database to determine a scene itself matching object under a capture boundary of the all-around perception scene, including:

[0024] Based on the current position of the open-pit truck, a preset scene consistent with the current position is obtained from a position-scene mapping table, and a preset sub-database consistent with the preset scene is called from the preset database;

[0025] Based on the all-around perception scene, a closed contour other than the truck contour of the open-pit truck is obtained as a first contour;

[0026] Each first contour is matched with each preset contour in the preset sub-database in turn to obtain a first matching result;

[0027] If the number of matching successful contours in the first matching result is greater than or equal to a preset number of matches, the matching successful first scene object and a second contour set that is not matched successfully are obtained;

[0028] Based on each contour edge of each second contour in the second contour set and a preset contour analysis width, a capture boundary of each contour edge is obtained;

[0029] Each capture boundary is unit boundary split to determine the color value of the point cloud originally contained in each boundary unit, and the average color value of the corresponding boundary unit is obtained by averaging, and a first color value map of the corresponding capture boundary is obtained;

[0030] The chromaticity difference under the same light intensity caused by the environmental pollution corresponding to the all-around perception scene at the current moment is obtained to obtain a corresponding color value influence factor;

[0031] Each first color value map and the color value influence factor are input into a color value influence removal model to obtain a second color value map with corresponding color value restoration;

[0032] The first distance of each second point cloud of each second color value map from the perception system of the corresponding open-pit truck is determined;

[0033] According to a weight setting rule, a second color value weight of each second point cloud in the second color value map is set;

[0034] In the second color value map, the second point cloud with a second color value weight less than a preset color value weight is removed to obtain a third color value map, and the point cloud corresponding to the highest color value in the third color value map is locked;

[0035] The coordinate system of the third color value map is constructed with the locked point cloud as the origin, and the transverse and longitudinal partial derivatives of each third point cloud in the third color value map based on the coordinate system are obtained;

[0036] Construct a horizontal deflection fitting graph and a vertical deflection fitting graph based on all horizontal deflection derivatives and vertical deflection derivatives involved in the same third color value graph;

[0037] Calculate a first numerical difference between two adjacent fitting points in the horizontal deflection fitting graph, and remove the horizontal deflection fitting points with a first numerical difference greater than a first number in the preset horizontal deflection numerical difference, and calculate a second numerical difference between two adjacent fitting points in the vertical deflection fitting graph, and remove the vertical deflection fitting points with a second numerical difference greater than a second number in the preset vertical deflection numerical difference, to obtain a fourth color value graph;

[0038] Take the contour of each fourth color value graph as a third contour, and re-match it with the preset sub-library to obtain the corresponding scene object, and obtain a second matching result;

[0039] The first matching result and the second matching result are taken as a scene itself matching object.

[0040] Preferably, the present application provides a perception system optimization method for an open-pit mine truck, each first contour is matched with each preset contour in the preset sub-library in turn to obtain a first matching result, comprising:

[0041] If the number of contours matched successfully in the first matching result is less than the preset number of matches, the adjacent position segment of the current position is locked according to the current driving state of the same open-pit mine truck;

[0042] The scene consistent with each position point in the adjacent position segment and the corresponding first contour are re-matched in turn.

[0043] Preferably, the present application provides a perception system optimization method for an open-pit mine truck, the scene itself matching object is evaluated, and if the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized, comprising:

[0044] An evaluation index is calculated according to the first scene object and the second scene object contained in the scene itself matching object;

[0045] If the evaluation index is greater than the standard evaluation index in the standard evaluation condition, it is determined that the perception system does not need to be optimized.

[0046] Preferably, the present application provides a perception system optimization method for an open-pit mine truck, an evaluation index is calculated according to the first scene object and the second scene object contained in the scene itself matching object, comprising:

[0047] ; wherein, represents the perception area of the i1th first scene object; represents the standard area of the i1th first scene object; represents the object weight of the ith1 first scene object; n1 represents the number of first scene objects contained in the scene itself matching object; n2 represents the number of second scene objects contained in the scene itself matching object; represents the perceived area of the ith2 second scene object; represents the standard area of the ith2 second scene object; represents the object weight of the ith2 second scene object; represents the total number of objects involved in the preset scene matched by the open-pit truck at the current position; represents the non-appeared scene object; represents the corresponding evaluation index.

[0048] Preferably, the present application provides a perception system optimization method for an open-pit truck, determining the optimization items of the perception system to be optimized, comprising:

[0049] determining the non-appeared scene object and the object features of the non-appeared scene object in all objects of the whole-body perception scene except the scene itself matching object;

[0050] obtaining the perception optimization factor consistent with the object features based on the feature-optimization database;

[0051] optimizing the optimization items of the perception system based on the perception optimization factor.

[0052] Other features and advantages of the present application will be described in the following description, and some will become apparent from the description, or will be learned through practice of the present application. The objects and other advantages of the present application will be achieved and obtained by the structure particularly pointed out in the written description, claims, and drawings.

[0053] The technical solutions of the present application will be further described in detail below with the help of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0054] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application, and are used to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0055] Figure 1 is a flow chart of a perception system optimization method for an open-pit truck in an embodiment of the present application. DETAILED DESCRIPTION

[0056] The preferred embodiments of the present application will be described below in conjunction with the drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not constitute a limitation on the present application.

[0057] Embodiment 1:

[0058] The embodiment of the present application provides a perception system optimization method for an open-pit truck, as shown in the figure, comprising: Figure 1

[0059] Step 1: based on the perception system arranged on the open-pit truck, the surrounding information in the driving range of the same open-pit truck is captured in real time;

[0060] Step 2: based on the real-time perception capture result, the surrounding perception scene of the same open-pit truck is constructed;

[0061] Step 3: the surrounding perception scene is matched with a preset database to determine the scene itself matching object under the capture boundary of the surrounding perception scene;

[0062] Step 4: the scene itself matching object is evaluated, if the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized;

[0063] Otherwise, the optimization item of the perception system is determined and optimized.

[0064] In the embodiment, the perception system refers to a system that accurately perceives the surrounding environment of the automatic driving vehicle by taking the data of multiple sensors and the information of the high-precision map as input, and performing a series of calculations and processing.

[0065] In the embodiment, the surrounding information refers to the information obtained by the perception system of the open-pit truck in the driving range, wherein the driving range refers to a range obtained according to the width required for normal driving of the open-pit truck and the length required for avoiding obstacles.

[0066] In the embodiment, real-time perception capture refers to real-time acquisition of the surrounding information in the driving range by the perception system of the open-pit truck.

[0067] In the embodiment, the real-time perception capture result refers to the point cloud data of each space point of the surrounding space of the open-pit truck, wherein the surrounding space is a three-dimensional space in the driving range, the space point refers to a point in the three-dimensional space, and the point cloud data refers to data information including three-dimensional coordinates X, Y, Z, color, classification value, intensity value and time.

[0068] In the embodiment, the surrounding perception scene refers to a three-dimensional image of the open-pit truck and a three-dimensional image of the surrounding space of the open-pit truck.

[0069] ​In the embodiment, the preset database refers to a database including preset scenes and corresponding preset sub-databases, wherein the preset sub-database refers to a database pre-set to contain three-dimensional images of instrument objects and article objects in the preset scene.

[0070] In the embodiment, the capture boundary refers to a boundary of each contour edge of the second contour captured according to a preset contour analysis width, wherein the preset contour analysis width refers to a width pre-set for contour analysis.

[0071] In the embodiment, the scene itself matching object refers to an instrument object corresponding to a preset contour successfully matched by the whole-body perception scene through matching with instrument objects in the preset database.

[0072] In the embodiment, the evaluation result refers to an index capable of representing the accuracy of the evaluation matching calculated.

[0073] In the embodiment, the standard evaluation condition refers to a condition pre-set for evaluating the scene matching link of the perception system, including: a standard evaluation index.

[0074] In the embodiment, the to-be-optimized item refers to an instrument object corresponding to a three-dimensional contour that cannot be matched in the whole-body perception scene.

[0075] The working principle and beneficial effects of the above technical solution are as follows: through the perception system arranged on the surface mine car, the whole-body information in the driving range of the same surface mine car is captured in real time to construct the whole-body perception scene of the same surface mine car, the scene matching is performed with the preset database, the scene itself matching object under the capture boundary of the whole-body perception scene is determined for evaluation, if the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized, otherwise, the to-be-optimized item of the perception system is determined and optimized, the scene matching link of the perception system is accurately evaluated, the deficiency of the perception system is accurately found, and thus the perception system is improved and optimized, and the normal work of the surface mine car is ensured.

[0076] Embodiment 2:

[0077] According to the method provided in Embodiment 1, based on the perception system arranged on the surface mine car, the whole-body information in the driving range of the same surface mine car is captured in real time, including:

[0078] According to the mine car type of the surface mine car, the driving range of the surface mine car is determined;

[0079] The current position of the surface mine car is positioned in real time, and when the positioning result overlaps with the driving range, the perception system on the surface mine car is controlled to start working;

[0080] Based on the perception system, the real-time perception capture is performed on the all-around information of the strip mine vehicle, wherein the real-time perception capture result is point cloud data of each spatial point of the all-around space of the strip mine vehicle.

[0081] In this embodiment, the type of the strip mine vehicle refers to the type according to the loadable tonnage of the strip mine vehicle, including: large mine vehicle, medium mine vehicle and small mine vehicle.

[0082] In this embodiment, the driving range refers to a range obtained according to the width required for normal driving of the strip mine vehicle and the length required for avoiding obstacles, which is set in advance according to the type of the strip mine vehicle.

[0083] In this embodiment, the perception system refers to a system that accurately perceives the surrounding environment of the autonomous vehicle by taking the data of multiple sensors and the information of the high-precision map as input, and performing a series of calculations and processing.

[0084] In this embodiment, the all-around information refers to the information obtained by the perception system of the strip mine vehicle within the driving range.

[0085] In this embodiment, the real-time perception capture result refers to point cloud data of each spatial point of the all-around space of the strip mine vehicle, wherein the all-around space is a three-dimensional space within the driving range, the spatial point refers to a point in the three-dimensional space, and the point cloud data refers to data information including three-dimensional coordinates X, Y, Z, color, classification value, intensity value and time.

[0086] In this embodiment, the spatial point refers to a point in the three-dimensional space.

[0087] In this embodiment, the point cloud data refers to data information including three-dimensional coordinates X, Y, Z, color, classification value, intensity value and time.

[0088] The working principle and beneficial effects of the above technical solution are: by determining the driving range of the strip mine vehicle, when the positioning result overlaps with the driving range, the perception system is controlled to perform real-time perception capture on the all-around information of the strip mine vehicle, accurate capture of real-time information is beneficial to accurate optimization of the perception system.

[0089] Embodiment 3:

[0090] According to the method provided in embodiment 1 of the application, in the process of constructing the all-around perception scene of the same strip mine vehicle based on the real-time perception capture result, the following steps are included:

[0091] The real-time perception capture result of the same strip mine vehicle is extracted in the same longitudinal coordinate layer.

[0092] Determine the gray scale difference absolute value of adjacent point clouds in each point cloud layer, and when the gray scale difference absolute value is greater than a preset edge difference value, lock the final point cloud according to max(x01, x02), wherein x01 represents a first point cloud in the adjacent point clouds; x02 represents a second point cloud in the adjacent point clouds; and max represents a maximum value symbol;

[0093] Based on all the final point clouds of the same point cloud layer, obtain a layer contour;

[0094] Combine all the layer contours in the order of the longitudinal coordinate to construct a three-dimensional scene of the same open-pit truck.

[0095] In this embodiment, the layer extraction of the same longitudinal coordinate means that the point cloud of the same longitudinal coordinate is taken as a layer, and all the point clouds of the layer are extracted.

[0096] In this embodiment, the point cloud layer refers to a set of point clouds of the same longitudinal coordinate.

[0097] In this embodiment, the preset edge difference value refers to a preset minimum value of the gray scale difference absolute value of the adjacent point clouds of the contour edge, and when the gray scale difference absolute value is greater than the preset edge difference value, the larger point cloud in the corresponding adjacent point clouds is the contour edge.

[0098] In this embodiment, the final point cloud refers to the larger point cloud in the adjacent point clouds.

[0099] In this embodiment, the layer contour refers to a contour obtained by connecting all the final point clouds of the same point cloud layer.

[0100] In this embodiment, the three-dimensional scene refers to a scene representing a three-dimensional image in a driving range obtained by combining all the layer contours in the order of the longitudinal coordinate.

[0101] The working principle and beneficial effects of the above technical solution are: by performing contour analysis on the point cloud contour in the order of the longitudinal coordinate, a three-dimensional scene is more accurately constructed.

[0102] Embodiment 4:

[0103] According to the method provided in Embodiment 1, based on the real-time perception capture result, a whole-body perception scene of the same open-pit truck is constructed, including:

[0104] Based on the vehicle structure parameters of the same open-pit truck, a three-dimensional truck of the same open-pit truck is obtained.

[0105] Based on the three-dimensional truck and the corresponding three-dimensional scene, a whole-body perception scene of the same open-pit truck is constructed.

[0106] In this embodiment, the vehicle structure parameters refer to the length, width and height of the vehicle and the basic structure form of the vehicle.

[0107] In this embodiment, the three-dimensional mine car refers to the three-dimensional image of the mine car obtained by reproducing the structural parameters of the vehicle.

[0108] In this embodiment, the whole-body perception scene refers to the three-dimensional image of the mine car and the three-dimensional image of the whole-body space of the mine car combined to obtain the three-dimensional image of the scene around the mine car.

[0109] The working principle and beneficial effects of the above technical solution are: by obtaining the three-dimensional image of the mine car, and combining the three-dimensional image of the whole-body space to obtain the three-dimensional image of the scene around the mine car, it is beneficial to the analysis of the information around the mine car.

[0110] Embodiment 5:

[0111] According to the method provided in Embodiment 1 of the application, the whole-body perception scene is matched with the preset database to determine the scene itself matching object under the capture boundary of the whole-body perception scene, which includes:

[0112] Based on the current position of the mine car, a preset scene consistent with the current position is obtained from a position-scene mapping table, and a preset sub-database consistent with the preset scene is called from the preset database;

[0113] Based on the whole-body perception scene, an enclosed contour other than the mine car contour of the mine car is obtained, and is taken as a first contour;

[0114] Each first contour is matched with each preset contour in the preset sub-database in turn to obtain a first matching result;

[0115] If the number of contours that match successfully in the first matching result is greater than or equal to a preset number of matches, the first scene object that matches successfully and the second contour set that does not match successfully are obtained;

[0116] Based on each contour edge of each second contour in the second contour set and the preset contour analysis width, the capture boundary of each contour edge is obtained;

[0117] Each capture boundary is unit boundary split to determine the color value of the point cloud originally contained in each boundary unit, and the average color value of the corresponding boundary unit is obtained by averaging, and a first color value map of the corresponding capture boundary is obtained;

[0118] The chroma difference under the same light intensity caused by the environmental pollution corresponding to the current moment of the whole-body perception scene is obtained, and the corresponding color value influence factor is obtained;

[0119] Each first color value map and the color value influence factor are input into a color value influence removal model to obtain a second color value map with corresponding color value restoration;

[0120] determining a first distance of each second point cloud of each second color value map from a perception system of a corresponding surface mine vehicle;

[0121] setting a second color value weight for each second point cloud in the second color value map according to a weight setting rule;

[0122] removing, in the second color value map, a second point cloud with a second color value weight less than a preset color value weight, obtaining a third color value map, and locking a point cloud corresponding to a highest color value in the third color value map;

[0123] constructing a coordinate system of the third color value map with the locked point cloud as an origin, and obtaining a transverse partial derivative and a longitudinal partial derivative of each third point cloud in the third color value map based on the coordinate system;

[0124] constructing a transverse partial fitting map and a longitudinal partial fitting map based on all transverse partial derivatives and longitudinal partial derivatives involved in a same third color value map;

[0125] calculating a first numerical difference between two adjacent fitting points in the transverse partial fitting map, removing a transverse partial fitting point with a first number of a preset transverse partial numerical difference greater than the first numerical difference, and calculating a second numerical difference between two adjacent fitting points in the longitudinal partial fitting map, removing a longitudinal partial fitting point with a second number of a preset longitudinal partial numerical difference greater than the second numerical difference, to obtain a fourth color value map;

[0126] taking an outline of each fourth color value map as a third outline, and re-matching the third outline with the preset sub-database to obtain a corresponding scene object, to obtain a second matching result;

[0127] wherein the first matching result and the second matching result are taken as a scene itself matching object.

[0128] In this embodiment, the position-scene mapping table refers to a mapping table in which positions and corresponding scenes are in one-to-one correspondence.

[0129] In this embodiment, the preset scene refers to a preset scene three-dimensional image obtained by dividing a surface mine according to positions.

[0130] In this embodiment, the preset database refers to a database including a preset scene and a corresponding preset sub-database, wherein the preset sub-database refers to a database including three-dimensional images of instrument objects and article objects in the preset scene.

[0131] In this embodiment, the preset sub-database refers to a database including three-dimensional images of instrument objects and article objects in the preset scene.

[0132] In this embodiment, the first outline refers to a closed outline excluding a mine vehicle outline of the surface mine vehicle in the surrounding perception scene.

[0133] In this embodiment, the preset contour refers to the contour of the three-dimensional image of the instrument object, the article object in the preset scene in the preset sub-library.

[0134] In this embodiment, the first matching result refers to the matching successful first contour and the corresponding preset contour obtained by matching each first contour with each preset contour.

[0135] In this embodiment, the preset number of matches refers to the minimum number of matching successful first contours that are preset to represent that the preset scene is the same as the whole-body perception scene.

[0136] In this embodiment, the first scene object refers to the corresponding object of the matching successful preset contour of the first contour.

[0137] In this embodiment, the second contour set refers to the closed contour in the whole-body perception scene except for the mine car contour of the surface mine car.

[0138] In this embodiment, the preset contour analysis width refers to the width required for preset contour analysis.

[0139] In this embodiment, the capture boundary refers to the boundary of each contour edge of the second contour taken according to the preset contour analysis width.

[0140] In this embodiment, the unit boundary refers to the minimum unit of preset boundary analysis, which contains a plurality of point clouds.

[0141] In this embodiment, the first color value map refers to a color value map of the capture boundary, which is composed of the average value of the color values of the point clouds in all boundary units of the capture boundary.

[0142] In this embodiment, the color value influence factor refers to the degree of color value influence under the same illumination brightness caused by pollution of the environment at the corresponding moment.

[0143] In this embodiment, the color value influence removal model refers to a model trained by the color value map and the color value influence factor to remove the influence of the color value influence factor.

[0144] In this embodiment, the second color value map refers to a color value map obtained by removing the influence of the color value influence factor on the first color value map through the color value influence removal model.

[0145] In this embodiment, the second point cloud refers to the point cloud in the second color value map.

[0146] In this embodiment, the first distance refers to the distance between each second point cloud and the perception system of the corresponding surface mine car.

[0147] In this embodiment, the weight setting rule is a rule of setting a weight for each second point cloud according to the first distance, wherein the smaller the first distance is, the higher the weight setting is.

[0148] In this embodiment, the second color value weight refers to the weight set for each second point cloud according to the weight setting rule and the first distance.

[0149] In this embodiment, the preset color value weight refers to a minimum value of a color value weight with reference value that is set in advance.

[0150] In this embodiment, the third color value map refers to a map of color values of a point cloud with reference value obtained by removing the second point cloud with a second color value weight less than the preset color value weight.

[0151] In this embodiment, the third point cloud refers to a point cloud in the third color value map.

[0152] In this embodiment, the horizontal partial derivative refers to a partial derivative of a point in the third color value map with respect to the horizontal axis.

[0153] In this embodiment, the vertical partial derivative refers to a partial derivative of a point in the third color value map with respect to the vertical axis.

[0154] In this embodiment, the horizontal partial fitting map refers to a fitting map of the horizontal partial derivative of each point cloud in the third color value map.

[0155] In this embodiment, the vertical partial fitting map refers to a fitting map of the vertical partial derivative of each point cloud in the third color value map.

[0156] In this embodiment, the first numerical difference refers to a difference between numerical values of the horizontal partial derivative corresponding to two adjacent fitting points in the horizontal partial fitting map.

[0157] In this embodiment, the preset horizontal partial numerical difference refers to a minimum value of a horizontal partial numerical difference with reference value that is set in advance.

[0158] In this embodiment, the second numerical difference refers to a difference between numerical values of the vertical partial derivative corresponding to two adjacent fitting points in the vertical partial fitting map.

[0159] In this embodiment, the preset vertical partial numerical difference refers to a minimum value of a vertical partial numerical difference with reference value that is set in advance.

[0160] In this embodiment, the fourth color value map refers to a color value map obtained by removing the horizontal partial fitting point and the vertical partial fitting point without reference value.

[0161] In this embodiment, the third contour refers to a contour of the fourth color value map.

[0162] In this embodiment, the second matching result refers to the matching successful third contour and the corresponding preset contour obtained by matching the processed third contour with each preset contour.

[0163] In this embodiment, the scene itself matching object refers to the matching successful preset contour corresponding to the first matching result and the second matching result.

[0164] The working principle and beneficial effects of the above technical solution are: by matching the closed contour in the surrounding perception scene and the preset sub-library corresponding to the current position of the surface mine car, the contours that do not pass the scene matching are treated, the point cloud color value with reference value is obtained, secondary matching is performed, the accuracy of scene matching is improved, the shortcomings of the scene matching stage of the perception system are optimized, and the normal work of the surface mine car is ensured.

[0165] Embodiment 6:

[0166] According to the method provided in embodiment 1 of the application, each first contour is matched with each preset contour in the preset sub-library in turn, and a first matching result is obtained, including:

[0167] If the number of matching successful contours in the first matching result is less than the preset number of matches, then according to the current driving state of the same surface mine car, the adjacent position segment of the current position is locked;

[0168] The scene consistent with each position point in the adjacent position segment is re-matched with the corresponding first contour in turn.

[0169] In this embodiment, the current driving state refers to the loading state, driving state or unloading state obtained by analyzing the state monitoring of the same surface mine car.

[0170] In this embodiment, the adjacent position segment refers to the position segment adjacent to the position segment where the current position is located.

[0171] The working principle and beneficial effects of the above technical solution are: by matching the contours that do not pass the scene matching with the scene, the accuracy of the scene matching of the perception system is improved.

[0172] Embodiment 7:

[0173] According to the method provided in embodiment 1 of the application, the scene itself matching object is evaluated, and if the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized, including:

[0174] According to the first scene object and the second scene object contained in the scene itself matching object, an evaluation index is calculated;

[0175] If the evaluation index is greater than a standard evaluation index in a standard evaluation condition, it is determined that the perception system does not need to be optimized.

[0176] In this embodiment, the evaluation index refers to an index that can represent the accuracy of the evaluation matching obtained by calculation.

[0177] In this embodiment, the standard evaluation condition refers to a condition for evaluating the scene matching link of the perception system that is set in advance, and includes a standard evaluation index.

[0178] In this embodiment, the standard evaluation index refers to a standard index that can represent the accuracy of the evaluation matching.

[0179] The working principle and beneficial effects of the above technical solution are: by calculating the parameters of the first scene object and the second scene object, the evaluation index is obtained, the scene matching link of the perception system is accurately evaluated, and the deficiencies of the perception system are accurately found.

[0180] Embodiment 8:

[0181] According to the method provided in Embodiment 1 of the application, the evaluation index is calculated according to the first scene object and the second scene object contained in the scene itself matching object, and includes:

[0182] ; wherein, represents the perception area of the i1th first scene object; represents the standard area of the i1th first scene object; represents the object weight of the i1th first scene object; n1 represents the number of first scene objects contained in the scene itself matching object; n2 represents the number of second scene objects contained in the scene itself matching object; represents the perception area of the i2th second scene object; represents the standard area of the i2th second scene object; represents the object weight of the i2th second scene object; represents the total number of objects involved in the preset scene matched by the open-pit truck at the current position; represents that the scene object does not exist; represents the corresponding evaluation index.

[0183] The working principle and beneficial effects of the above technical solution are: by calculating the parameters of the first scene object and the second scene object, the evaluation index is obtained, the scene matching link of the perception system is accurately evaluated, and the deficiencies of the perception system are accurately found.

[0184] Embodiment 9:

[0185] According to the method provided by the embodiment 1 of the application, the items to be optimized of the perception system are determined to be optimized, comprising:

[0186] The non-appeared scene object and the object feature of the non-appeared scene object are determined from all objects of the whole-body perception scene except the matched object of the scene itself;

[0187] The perception optimization factor consistent with the object feature is obtained based on the feature-optimization database;

[0188] The items to be optimized of the perception system are optimized based on the perception optimization factor.

[0189] In this embodiment, the non-appeared scene object refers to the instrument object corresponding to the three-dimensional contour which cannot be matched in the whole-body perception scene.

[0190] In this embodiment, the object feature refers to the feature of the non-appeared scene object, comprising: three-dimensional contour, three-dimensional color value map of the non-appeared object, and use.

[0191] In this embodiment, the feature-optimization database refers to the database containing the object feature and the perception optimization factor corresponding thereto.

[0192] In this embodiment, the perception optimization factor refers to the three-dimensional modeling for optimizing the perception system converted from the object feature.

[0193] The working principle and beneficial effects of the above technical solution are: by analyzing the object feature of the non-appeared scene object, the corresponding perception optimization factor is obtained, the instrument object not in the preset database of the perception system is added, the perception system is optimized, and the normal work of the open-pit mine car is ensured.

[0194] Obviously, those skilled in the art can make various modifications and variations to the application without departing from the spirit and scope of the application. Thus, if these modifications and variations of the application belong to the scope of the claims of the application and the equivalent technology thereof, the application also intends to include these modifications and variations.

Claims

1. A method for perception system optimization for a surface mine truck, the method comprising: The method comprises the following steps: Step 1: based on the sensing system arranged on the open-pit truck, the surrounding information within the driving range of the same open-pit truck is captured in real time; Step 2: based on the real-time sensing capture result, the surrounding sensing scene of the same open-pit truck is constructed; Step 3: the surrounding sensing scene is matched with the preset database to determine the scene itself matching object under the capture boundary of the surrounding sensing scene, specifically comprising: Based on the current position of the open-pit truck, the preset scene consistent with the current position is obtained from the position-scene mapping table, and the preset sub-database consistent with the preset scene is called from the preset database; Based on the surrounding sensing scene, an enclosed contour except the truck contour of the open-pit truck is obtained as a first contour; Each first contour is matched with each preset contour in the preset sub-database in turn to obtain a first matching result; If the number of matching successful contours in the first matching result is greater than or equal to a preset number of matches, the first scene object matching successfully and a second contour set matching unsuccessfully are obtained; Based on each contour edge of each second contour in the second contour set and the preset contour analysis width, the capture boundary of each contour edge is obtained; Each capture boundary is unit boundary split to determine the color value of the point cloud originally contained in each boundary unit, and the average color value of the corresponding boundary unit is obtained by averaging, and a first color value map of the corresponding capture boundary is obtained; The chromaticity difference under the same light intensity caused by the environmental pollution corresponding to the current moment of the surrounding sensing scene is obtained, and the corresponding color value influence factor is obtained; Each first color value map and the color value influence factor are input into the color value influence removal model to obtain a second color value map with corresponding color value restoration; The first distance of each second point cloud of each second color value map from the sensing system of the corresponding open-pit truck is determined; According to the weight setting rule, the second color value weight of each second point cloud in the second color value map is set; In the second color value map, the second point cloud with a second color value weight less than a preset color value weight is removed to obtain a third color value map, and the point cloud corresponding to the highest color value in the third color value map is locked; Taking the locked point cloud as the origin, a coordinate system of the third color value map is constructed, and the transverse and longitudinal partial derivatives of each third point cloud in the third color value map based on the coordinate system are obtained; Based on all the transverse and longitudinal partial derivatives involved in the same third color value map, a transverse fitting graph and a longitudinal fitting graph are constructed respectively; The first numerical difference between two adjacent fitting points in the transverse fitting graph is calculated, and the transverse fitting points with a first numerical difference greater than a first number of a preset transverse numerical difference are removed, and the second numerical difference between two adjacent fitting points in the longitudinal fitting graph is calculated, and the longitudinal fitting points with a second numerical difference greater than a second number of a preset longitudinal numerical difference are removed to obtain a fourth color value map; The contour of each fourth color value map is taken as a third contour, and re-matching is performed with the preset sub-database to obtain the corresponding scene object, and a second matching result is obtained; The first matching result and the second matching result are taken as the scene itself matching object. Step 4: evaluating the object matching the scene itself, if the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized; Otherwise, determine the optimization items of the perception system and optimize.

2. The method of claim 1, wherein, Based on the perception system arranged on the open-pit truck, the surrounding information within the driving range of the same open-pit truck is captured in real time, including: According to the type of the open-pit truck, the driving range of the open-pit truck is determined; The current position of the open-pit truck is positioned in real time, and when the positioning result overlaps with the driving range, the perception system on the open-pit truck is controlled to start working; Based on the perception system, the surrounding information of the open-pit truck is captured in real time, wherein the real-time perception capture result is the point cloud data of each space point of the surrounding space of the open-pit truck.

3. The method of claim 1, wherein, Based on the real-time perception capture result, in the process of constructing the surrounding perception scene of the same open-pit truck, including: The real-time perception capture result of the same open-pit truck is extracted into the same longitudinal coordinate layer; The absolute value of the gray difference of adjacent point clouds in each point cloud layer is determined, and when the absolute value of the gray difference is greater than the preset edge difference, the final point cloud is locked according to max(x01, x02), wherein x01 represents the first point cloud in the adjacent point cloud; x02 represents the second point cloud in the adjacent point cloud; max represents the maximum value symbol; Based on all the final point clouds of the same point cloud layer, the layer contour is obtained; All layer contours are combined in order of longitudinal coordinate to construct the three-dimensional scene of the same open-pit truck.

4. The method of claim 3, wherein, Based on the real-time perception capture result, the surrounding perception scene of the same open-pit truck is constructed, including: Based on the vehicle structure parameters of the same open-pit truck, the three-dimensional truck of the same open-pit truck is obtained; Based on the three-dimensional truck and the corresponding three-dimensional scene, the surrounding perception scene of the same open-pit truck is constructed.

5. The method of claim 1, wherein, Each first contour is matched with each preset contour in the preset sub-library in turn to obtain a first matching result, including: If the number of contours that match successfully in the first matching result is less than the preset number of matches, the adjacent position segment of the current position is locked according to the current driving state of the same open-pit truck; The scene consistent with each position point in the adjacent position segment is re-matched with the corresponding first contour in turn.

6. The method of claim 1, wherein, The object matching the scene itself is evaluated, and if the evaluation result meets the standard evaluation condition, it is determined that the perception system does not need to be optimized, including: According to the first scene object and the second scene object contained in the object matching the scene itself, an evaluation index is calculated; If the evaluation index is greater than the standard evaluation index in the standard evaluation condition, it is determined that the perception system does not need to be optimized.

7. The method of claim 6, wherein, According to the first scene object and the second scene object contained in the object matching the scene itself, an evaluation index is calculated, including: ; wherein, represents the perceived area of the ith1first scene object; represents the standard area of the ith1first scene object; represents the object weight of the ith1first scene object; n1represents the number of first scene objects contained by the scene itself matching object; n2represents the number of second scene objects contained by the scene itself matching object; represents the perceived area of the ith2second scene object; represents the standard area of the ith2second scene object; represents the object weight of the ith2second scene object; represents the total number of objects involved in the preset scene matched by the open-pit truck at the current position; represents the non-occurrence of the scene object; represents the corresponding evaluation index.

8. The method of claim 1, wherein, Determine the optimization items of the perception system for optimization, including: Determine the object features of the non-appeared scene objects and the non-appeared scene objects in all objects of the surrounding perception scene except the object matching the scene itself; Based on the feature-optimization database, the perception optimization factor consistent with the object features is obtained; optimizing, based on the perception optimization factor, an item to be optimized of the perception system.

Images