CN116859413A - Perception model building method for open-air mine car - Google Patents
Perception model building method for open-air mine car Download PDFInfo
- Publication number
- CN116859413A CN116859413A CN202310641511.0A CN202310641511A CN116859413A CN 116859413 A CN116859413 A CN 116859413A CN 202310641511 A CN202310641511 A CN 202310641511A CN 116859413 A CN116859413 A CN 116859413A
- Authority
- CN
- China
- Prior art keywords
- road
- point cloud
- target
- mine car
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000008447 perception Effects 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000007781 pre-processing Methods 0.000 claims abstract description 22
- 238000013528 artificial neural network Methods 0.000 claims abstract description 20
- 230000008569 process Effects 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000011156 evaluation Methods 0.000 claims description 38
- 238000001514 detection method Methods 0.000 claims description 37
- 230000004888 barrier function Effects 0.000 claims description 30
- 238000001914 filtration Methods 0.000 claims description 25
- 230000003068 static effect Effects 0.000 claims description 19
- 238000012360 testing method Methods 0.000 claims description 17
- 238000005070 sampling Methods 0.000 claims description 16
- 230000004927 fusion Effects 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 13
- 230000000694 effects Effects 0.000 claims description 11
- 238000005065 mining Methods 0.000 claims description 11
- 230000033001 locomotion Effects 0.000 claims description 8
- 238000013135 deep learning Methods 0.000 claims description 7
- 239000013598 vector Substances 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 230000007613 environmental effect Effects 0.000 claims description 5
- 239000004744 fabric Substances 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000013095 identification testing Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 230000005484 gravity Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 description 9
- 238000012545 processing Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009897 systematic effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
- G01S17/931—Lidar systems specially adapted for specific applications for anti-collision purposes of land vehicles
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
- G01S7/4802—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Electromagnetism (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a perception model building method for an open-air mine car, which relates to the technical field of road perception and comprises the following steps: acquiring historical driving data acquired by utilizing a three-dimensional laser radar, and preprocessing the historical data to obtain target data; training a neural network based on the target data to obtain a road safety perception model; the road safety perception model is utilized to judge the condition of the road section in front of the open-air mine car in real time, and a first perception result is obtained and output; and the control center platform controls the mine car to respond according to the first sensing result. The neural network is trained to establish a road safety perception model based on preprocessing and analyzing the historical driving data acquired by utilizing the three-dimensional laser radar so as to effectively judge the condition of a road section in the driving process of the outdoor mine car, thereby controlling the mine car to automatically slow down or avoid and guaranteeing the driving safety.
Description
Technical Field
The invention relates to the technical field of road perception, in particular to a perception model building method for an open-air mine car.
Background
In recent years, emerging industries and technologies are rapidly developing, industrialization, mechanization and informatization of mining areas are also coming to break through the innovation, and mining engineering systems are increasingly huge and complex. And because the mine car transportation route is stable, the mine road is closed, the unmanned technique can be widely applied to the strip mine industry, the personnel participation is greatly reduced, the labor cost is saved, and the safety and the production efficiency of operators are improved. Meanwhile, because the mining area is remote, the road condition environment is complex, and the mine car production operation and the safety guarantee face great challenges.
Therefore, the invention provides a perception model building method for an open-air mine car vehicle, which is used for effectively perceiving the condition of a driving road section, so as to control the mine car to automatically slow down or avoid and ensure driving safety.
Disclosure of Invention
The invention provides a perception model building method for an open-air mine car vehicle, which is used for building a road safety perception model by training a neural network based on preprocessing and analyzing historical driving data acquired by utilizing a three-dimensional laser radar so as to realize effective judgment on the condition of a front road section in the driving process of the open-air mine car vehicle, thereby controlling the mine car to automatically slow down or avoid and guaranteeing driving safety.
The invention provides a perception model building method for an open-air mine car, which comprises the following steps:
step 1: acquiring historical driving data acquired by utilizing a three-dimensional laser radar, and preprocessing the historical driving data to obtain target data;
step 2: training a neural network based on the target data to obtain a road safety perception model;
step 3: the road safety perception model is utilized to judge the condition of the road section in front of the open-air mine car in real time, and a first perception result is obtained and output;
step 4: and the control center platform controls the mine car to respond according to the first sensing result.
Preferably, the method for obtaining the historical driving data acquired by the three-dimensional laser radar and preprocessing the historical driving data to obtain target data includes:
step 11: extracting a history point cloud frame acquired by scanning a front road by using a three-dimensional laser radar in the running process of a mine car with a preset amount from a mine database;
step 12: filtering dynamic target point clouds in historical point cloud frames through preprocessing, taking the first two frames in every three historical point cloud frames as historical frames, and taking the rest frames as current frames;
step 13: after the pose change matrixes of the historical frame and the current frame are determined, converting the point cloud data of the historical frame into a point cloud coordinate system of the current frame to obtain a first point cloud frame;
step 14: after the dynamic target point cloud clusters obtained in the preprocessing stage are clustered, the dynamic target point cloud in the first point cloud frame is offset eliminated by utilizing a dynamic target translation method, and a fusion frame is obtained and is output as target data.
Preferably, filtering the dynamic target point cloud of the historical point cloud frame includes:
step 21: projecting the point cloud data of the historical point cloud frames into a two-dimensional raster pattern,
step 22: according to the point cloud attribute in the grid, comparing the height difference between each point cloud and the lowest point in the grid with a first threshold value, and dividing the point cloud to obtain a ground point Yun Yufei ground point cloud;
step 23: filtering the ground point cloud, calculating and obtaining the maximum height difference between the rest points in the grids and the first number of the rest points in the grids, and regarding the corresponding grids as barrier grids if the maximum height difference is larger than a preset height threshold value and the first number is larger than a preset number threshold value;
step 24: considering the grids except the obstacle grid in the two-dimensional grid map as a drivable grid;
step 25: and determining a dynamic target point cloud cluster and a static target point cloud cluster from the obstacle point cloud grid according to the cluster characteristics of the obstacle point cloud clusters, and completing identification and filtering of the dynamic target point cloud.
Preferably, training the neural network to obtain the road safety perception model based on the target data includes:
step 31: taking a preset amount of historical driving data as an input party, and constructing a road safety perception neural network;
step 32: performing feature extraction on the target data by using a point cloud target detection algorithm based on deep learning to realize target category identification and endow corresponding candidate frames, so as to obtain a plurality of target frames;
step 33: acquiring the number and size parameters of the corresponding dynamic barriers and static barriers according to the target frame, and determining the barrier identification effect by combining the barrier identification test indexes;
step 34: and training a neural network in road safety perception by using three-quarter target frames as training samples and combining road flatness to obtain a road safety perception model, so as to calculate and obtain a road condition evaluation coefficient, wherein the formula is as follows:
the method comprises the steps of carrying out a first treatment on the surface of the In (1) the->The road condition evaluation coefficient corresponding to the ith target frame is expressed; />Road flatness expressed as a road section corresponding to the ith target frame; />The size of the area denoted as the ith obstacle in the ith target frame; />Denoted as the i-th target intra-frame road surface area; />The preset influence weight of the road flatness on road condition evaluation is expressed; />The preset influence weight of the specific gravity of the area occupied by the barrier-free area on the road surface on the road condition evaluation is expressed; />Expressed as a loss factor generated by environmental impact factors in the calculation of road condition assessment.
Preferably, according to the target frame, the number and size parameters of the corresponding dynamic obstacle and static obstacle are obtained, and the obstacle recognition effect is determined by combining the obstacle recognition test indexes, including:
randomly selecting a certain amount of target frames as test frames, combining and analyzing the number of dynamic obstacles and static obstacles in the test frames with the number of real obstacles, and obtaining the positive detection rate, the false detection rate and the omission rate of the corresponding obstacles by using a test index formula;
the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>A positive detection rate expressed as an obstacle in the jth target frame; />The correct number of detected obstacles in the jth target frame; />Expressed as the total number of obstacles in the jth target frame;a false detection rate expressed as an obstacle in the jth target frame; />A false detection number expressed as an obstacle in the jth target frame; />A number of obstacles represented as incorrectly classified in the jth target frame; />The false negative rate of the obstacle in the jth target frame; />Expressed as the total number of undetected obstacles in the jth target frame;
and according to the test index result, the false detection rate and the omission factor are both lower than the preset threshold, and the obstacle recognition effect is judged to be good.
Preferably, in the process of calculating the road condition evaluation coefficient, the method further includes: calculating the road flatness, specifically comprising:
step 41: extracting road points from the target data by using a cloth simulation algorithm, creating a track vector line along the track belt direction of the lane, and taking the road points in a preset range as corresponding point cloud data to be calculated;
step 42: acquiring the average value of the elevation in the preset neighborhood of the point cloud data to be calculated as the elevation value of the corresponding data of the point, and acquiring first point cloud data;
step 43: after segmenting the first point cloud data according to a preset calculation interval, acquiring the number of points in each section interval according to a preset sampling interval, so as to calculate and obtain an effective segmentation number;
step 44: the method comprises the steps of adopting a preset road slope as an initial value, solving by an iteration method to obtain 4 dynamic reaction quantities of the speed and the acceleration of the corresponding vehicle body spring bodies and non-spring bodies of the mine car running through all sampling points on the longitudinal and transverse surfaces, and then calculating to obtain an adjustment slope value of each sampling point;
step 45: and according to the gradient adjustment value, calculating to obtain a first flatness, and subtracting a system deviation obtained by comparing the first flatness with the calibrated flatness obtained by leveling calculation to obtain the road flatness.
Preferably, the real-time judgment is performed on the road situation of the road section in front of the outdoor mine car by using the road safety perception model, a first perception result is obtained and output, and the method comprises the following steps:
preprocessing point cloud frame data obtained by utilizing a three-dimensional laser radar installed on an outdoor mine car to acquire the condition of a road section in front of the vehicle in real time, and inputting the point cloud frame data into a road safety perception model to obtain a real-time road condition evaluation coefficient of the road in front;
if the real-time road condition evaluation coefficient is not smaller than a preset evaluation threshold, judging and calibrating the safety level of the front road to be three-level, and outputting the safety level as a first perception result;
if the real-time road condition evaluation coefficient is smaller than a preset evaluation threshold value and larger than a preset low threshold value, judging and calibrating the safety level of the front road as a second level, and outputting the second level as a first perception result;
otherwise, calibrating the safety level of the road in front as a first level and outputting the first level as a first sensing result.
Preferably, the control center station controls the mine car to respond according to the first sensing result, including:
the control center station collects and analyzes the first sensing result in real time, and if the safety level of the road in front is three-level, the mine car is controlled to normally run along a preset track according to the normal speed;
if the safety level of the road in front is two-level, calculating to obtain the running state of the corresponding dynamic obstacle based on RTK-GPS data and INS course angle data acquired in real time by the integrated navigation system, and predicting to obtain the corresponding target movement track based on Kalman filtering;
taking the areas except the static obstacle as a feasible area, and analyzing whether the mine car collides with the corresponding dynamic obstacle in the running process of the feasible area by combining the target movement track, and if so, predicting to obtain collision points;
connecting the dynamic barrier with the corresponding collision point position, changing the dynamic barrier into a forbidden area, and selecting a local sub-target node to conduct local obstacle avoidance path planning so as to ensure that the mine car avoids the dynamic barrier until reaching the target node;
if collision does not exist, controlling the mine car to run at a speed reduction in a drivable area when the real-time road flatness of the front road is smaller than a preset flatness threshold value;
if the front road safety level is one level, the control center console controls the mine car to brake emergently.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a flow chart of a method of creating a perception model for an open air vehicle in accordance with an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
The embodiment of the invention provides a perception model building method for an open-air mine car, as shown in fig. 1, comprising the following steps:
step 1: acquiring historical driving data acquired by utilizing a three-dimensional laser radar, and preprocessing the historical driving data to obtain target data;
step 2: training a neural network based on the target data to obtain a road safety perception model;
step 3: the road safety perception model is utilized to judge the condition of the road section in front of the open-air mine car in real time, and a first perception result is obtained and output;
step 4: and the control center platform controls the mine car to respond according to the first sensing result.
In the embodiment, the three-dimensional laser radar is deployed on a car body of a mine car and is used for transmitting detection signals to a target, then the received information reflected from the target is compared with the transmission signals, and related information of the target, such as parameters of the distance, the gesture, the shape and the like of the target, is obtained after proper processing, so that detection, tracking and identification of obstacles on a road are realized; historical driving data refers to data extracted from a mining database, such as historical point cloud frame data, acquired by scanning a road ahead by using a three-dimensional laser radar in the traveling process of a mine car; the target data refers to a fusion frame obtained by filtering dynamic target point clouds from historical point cloud frames in historical driving data, taking the first two frames in every three historical point cloud frames as historical frames, and fusing the rest frames as current frames.
In the embodiment, the road safety perception model is a perception model established based on a target data training neural network and is used for judging real-time road conditions in front of a mine car; the first perception result refers to a road security level obtained by analyzing the result output after the data acquired by the three-dimensional laser radar in real time are input into the road security perception model; the control center platform is used for controlling the mine car to realize automatic deceleration, detour or emergency braking after the first sensing result is collected and analyzed, so that driving safety is guaranteed.
The beneficial effects of the technical scheme are as follows: the neural network is trained to establish a road safety perception model based on the preprocessed historical driving data acquired by the three-dimensional laser radar, so that the condition of a road section in front of the surface mine car in the driving process is effectively judged, the mine car is controlled to automatically decelerate or avoid, and the driving safety is guaranteed.
The embodiment of the invention provides a perception model building method for an open-air mine car, which is used for acquiring historical driving data acquired by utilizing a three-dimensional laser radar and preprocessing the historical driving data to obtain target data, and comprises the following steps:
step 11: extracting a history point cloud frame acquired by scanning a front road by using a three-dimensional laser radar in the running process of a mine car with a preset amount from a mine database;
step 12: filtering dynamic target point clouds in historical point cloud frames through preprocessing, taking the first two frames in every three historical point cloud frames as historical frames, and taking the rest frames as current frames;
step 13: after the pose change matrixes of the historical frame and the current frame are determined, converting the point cloud data of the historical frame into a point cloud coordinate system of the current frame to obtain a first point cloud frame;
step 14: after the dynamic target point cloud clusters obtained in the preprocessing stage are clustered, the dynamic target point cloud in the first point cloud frame is offset eliminated by utilizing a dynamic target translation method, and a fusion frame is obtained and is output as target data.
In the embodiment, the mining database mainly comprises mining area data, such as road records, longitude and latitude position coordinates, mine car data, such as car body parameters, mine car running data, such as historical point cloud frames acquired by scanning a front road by using a three-dimensional laser radar in the running process, running speed and operator data; the preset amount is set in advance; the three-dimensional laser radar is used for detecting, tracking and identifying obstacles on a road; the point cloud frame is an aggregate image of sampling points with space coordinates, which is obtained by scanning a road in front of a mine car through a three-dimensional laser radar.
In the embodiment, preprocessing refers to projecting three-dimensional point cloud data to a two-dimensional grid graph, dividing an obstacle point cloud cluster according to point cloud attributes in the grid, and finding out a dynamic target point cloud cluster to realize rapid identification and filtering of the dynamic target point cloud; the purpose of filtering dynamic target point clouds in historical point cloud frames is to avoid accumulated errors caused by the dynamic target point clouds in the process of estimating a pose transformation matrix by point cloud registration of front and rear frames, so that positioning accuracy is improved, wherein the dynamic target point clouds refer to corresponding point cloud data of pedestrians and vehicles moving in a road section in front of a mine car, and the pose transformation matrix is used for realizing fusion of the first two frames in every three historical point cloud frames as historical frames and the rest frames as current frames; the first point cloud frame is obtained by converting historical frame point cloud data into a current frame point cloud coordinate system based on a displacement change matrix.
In this embodiment, the dynamic target point cloud cluster refers to a dynamic target point cloud data set; the dynamic target translation method refers to nearest neighbor searching of clustering results of point clouds of a historical frame and a current frame, and corresponding clustering targets in the historical frame are translated to positions of the targets in the current frame, so that the offset elimination problem is solved; the fusion frame is a frame image obtained by filtering dynamic target point clouds from historical point cloud frames in historical driving data, taking the first two frames in every three historical point cloud frames as historical frames, and taking the rest frames as current frames for fusion.
The beneficial effects of the technical scheme are as follows: the multi-frame point cloud fusion algorithm is utilized to process the historical point cloud frames to obtain fusion frames and output the fusion frames as target data, so that the problem of low positioning accuracy of the laser radar caused by moving pedestrians and vehicles in a real-time driving environment is effectively solved, and the accurate establishment of a follow-up model is facilitated.
The embodiment of the invention provides a perception model building method for an open-air mine car, which is used for filtering a historical point cloud frame dynamic target point cloud and comprises the following steps:
step 21: projecting the point cloud data of the historical point cloud frames into a two-dimensional raster pattern,
step 22: according to the point cloud attribute in the grid, comparing the height difference between each point cloud and the lowest point in the grid with a first threshold value, and dividing the point cloud to obtain a ground point Yun Yufei ground point cloud;
step 23: filtering the ground point cloud, calculating and obtaining the maximum height difference between the rest points in the grids and the first number of the rest points in the grids, and regarding the corresponding grids as barrier grids if the maximum height difference is larger than a preset height threshold value and the first number is larger than a preset number threshold value;
step 24: considering the grids except the obstacle grid in the two-dimensional grid map as a drivable grid;
step 25: and determining a dynamic target point cloud cluster and a static target point cloud cluster from the obstacle point cloud grid according to the cluster characteristics of the obstacle point cloud clusters, and completing identification and filtering of the dynamic target point cloud.
In the embodiment, the point cloud frame is an aggregate image of sampling points with space coordinates, which is obtained by scanning a road in front of a mine car through a three-dimensional laser radar; the point cloud data is a vector set in a three-dimensional coordinate system and comprises coordinate information and laser emission intensity information; the two-dimensional grid graph is based on the point cloud data of the historical point cloud frame to carry out grid division processing, so that the points in the grid can be judged to be ground points or obstacle points; the attributes of the point cloud mainly refer to surface normal vectors; the first threshold is set in advance; the ground point cloud refers to a set of sampling points with space coordinates on the road surface; the non-ground point cloud refers to a set of sampling points of non-road-surface-belt space coordinates, such as a vehicle point cloud and a pedestrian point cloud of a road.
In this embodiment, the preset height threshold value and the preset number threshold value are set in advance.
In this embodiment, for example, there are grids A1 and A2, the corresponding maximum height differences are both greater than a preset higher threshold, the first number of remaining points in the grid A1 is greater than a preset number threshold, and the first number of remaining points in the grid A2 is less than the preset number threshold, at which point the grid A1 is regarded as an obstacle grid.
In this embodiment, the drivable grid is a grid other than the obstacle grid in the two-dimensional raster pattern; the obstacle point cloud cluster refers to an obstacle point cloud data set; the cluster features are the number of points in the pointing cloud cluster, the sum of the horizontal coordinates of the points and the square sum of the horizontal coordinates; the dynamic target point cloud cluster refers to a set of dynamic target point cloud data, such as the point cloud data of moving pedestrians and the point cloud data of running vehicles; the static target point cloud cluster refers to a set of static target point cloud data, such as pothole point cloud data.
The beneficial effects of the technical scheme are as follows: after the point cloud data of the historical point cloud frames are projected to the two-dimensional grid graph to filter the ground point cloud, the height difference and the number of the point clouds are utilized to judge the barrier grids, and effective identification and filtering of the dynamic target point cloud are completed under the conditions of eliminating the ground point influence and reducing the environmental noise interference.
The embodiment of the invention provides a perception model building method for an open-air mine car, which is used for training a neural network to obtain a road safety perception model based on target data, and comprises the following steps:
step 31: taking a preset amount of historical driving data as an input party, and constructing a road safety perception neural network;
step 32: performing feature extraction on the target data by using a point cloud target detection algorithm based on deep learning to realize target category identification and endow corresponding candidate frames, so as to obtain a plurality of target frames;
step 33: acquiring the number and size parameters of the corresponding dynamic barriers and static barriers according to the target frame, and determining the barrier identification effect by combining the barrier identification test indexes;
step 34: and training a neural network in road safety perception by using three-quarter target frames as training samples and combining road flatness to obtain a road safety perception model, so as to calculate and obtain a road condition evaluation coefficient, wherein the formula is as follows:
the method comprises the steps of carrying out a first treatment on the surface of the In (1) the->The road condition evaluation coefficient corresponding to the ith target frame is expressed; />Road flatness expressed as a road section corresponding to the ith target frame; />The size of the area denoted as the ith obstacle in the ith target frame; />Denoted as the i-th target intra-frame road surface area; />The preset influence weight of the road flatness on road condition evaluation is expressed; />The preset influence weight of the specific gravity of the area occupied by the barrier-free area on the road surface on the road condition evaluation is expressed; />Expressed as a loss factor generated by environmental impact factors in the calculation of road condition assessment.
In this embodiment, the preset amount is set in advance; historical driving data refers to data extracted from a mining database, such as historical point cloud frame data, acquired by scanning a road ahead by using a three-dimensional laser radar in the traveling process of a mine car; the point cloud target detection algorithm based on deep learning is used for realizing target identification and candidate frame positioning after feature learning and feature extraction are performed on target data.
In this embodiment, the target data refers to a fusion frame obtained by filtering dynamic target point clouds from historical point cloud frames in historical driving data, taking the first two frames in every three historical point cloud frames as historical frames, and taking the remaining frames as current frames for fusion; the target categories mainly include dynamic and static obstacles, wherein the static obstacles are classified into convex obstacles, such as mining tools, and concave obstacles, such as potholes; dynamic obstacles generally refer to moving pedestrians and traveling vehicles; the target frame refers to a frame image obtained after the target data, namely the fusion frame features, are extracted by a point cloud target detection algorithm based on deep learning, and the target category is identified and assigned with a corresponding candidate frame.
In this embodiment, the obstacle recognition effect is determined by an obstacle recognition test index, wherein the obstacle recognition test index includes a positive detection rate, a false detection rate, and a false detection rate of the obstacle; road flatness is one of the main technical indexes for evaluating the road surface quality of a road; the road safety perception model is a perception model established by combining road flatness and training a neural network based on target data, and is used for judging real-time road conditions in front of the mine car; the road condition evaluation coefficient is used for judging the road safety level; the environmental impact factors mainly refer to weather and traffic flow factors.
The beneficial effects of the technical scheme are as follows: the target frame training obtained by processing the target data by utilizing the point cloud target detection algorithm based on deep learning takes the historical driving data as an input party, and a neural network for obtaining road safety perception is constructed to effectively establish a road safety perception model, accurately judge the real-time road condition in front of the mine car, respond in time and guarantee the travelling safety.
The embodiment of the invention provides a perception model building method for an open-air mine car, which is used for acquiring the number and size parameters of corresponding dynamic barriers and static barriers according to a target frame, and determining the barrier identification effect by combining barrier identification test indexes, and comprises the following steps:
randomly selecting a certain amount of target frames as test frames, combining and analyzing the number of dynamic obstacles and static obstacles in the test frames with the number of real obstacles, and obtaining the positive detection rate, the false detection rate and the omission rate of the corresponding obstacles by using a test index formula;
the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>A positive detection rate expressed as an obstacle in the jth target frame; />The correct number of detected obstacles in the jth target frame; />Expressed as the total number of obstacles in the jth target frame;a false detection rate expressed as an obstacle in the jth target frame; />A false detection number expressed as an obstacle in the jth target frame; />A number of obstacles represented as incorrectly classified in the jth target frame; />The false negative rate of the obstacle in the jth target frame; />Expressed as the total number of undetected obstacles in the jth target frame;
and according to the test index result, the false detection rate and the omission factor are both lower than the preset threshold, and the obstacle recognition effect is judged to be good.
In this embodiment, the test index result refers to the positive detection rate, the false detection rate, and the omission rate of the obstacle; the preset threshold is set in advance based on the mine road complexity.
The beneficial effects of the technical scheme are as follows: the positive detection rate, the false detection rate and the omission rate of the obstacle are calculated by utilizing a formula, and the recognition effect of the point cloud target detection algorithm based on deep learning on the obstacle is determined to be good, so that the method has a certain feasibility for identifying the obstacle.
The embodiment of the invention provides a perception model building method for an open-air mine car, which further comprises the following steps in the process of calculating road condition evaluation coefficients: calculating the road flatness, specifically comprising:
step 41: extracting road points from the target data by using a cloth simulation algorithm, creating a track vector line along the track belt direction of the lane, and taking the road points in a preset range as corresponding point cloud data to be calculated;
step 42: acquiring the average value of the elevation in the preset neighborhood of the point cloud data to be calculated as the elevation value of the corresponding data of the point, and acquiring first point cloud data;
step 43: after segmenting the first point cloud data according to a preset calculation interval, acquiring the number of points in each section interval according to a preset sampling interval, so as to calculate and obtain an effective segmentation number;
step 44: the method comprises the steps of adopting a preset road slope as an initial value, solving by an iteration method to obtain 4 dynamic reaction quantities of the speed and the acceleration of the corresponding vehicle body spring bodies and non-spring bodies of the mine car running through all sampling points on the longitudinal and transverse surfaces, and then calculating to obtain an adjustment slope value of each sampling point;
step 45: and according to the gradient adjustment value, calculating to obtain a first flatness, and subtracting a system deviation obtained by comparing the first flatness with the calibrated flatness obtained by leveling calculation to obtain the road flatness.
In the embodiment, the cloth simulation algorithm is used for filtering out points with invalid road flatness calculation to obtain road points; the target data refers to a fusion frame obtained by filtering dynamic target point clouds from historical point cloud frames in historical driving data, taking the first two frames in every three historical point cloud frames as historical frames and taking the rest frames as current frames for fusion; the track vector line is created based on the direction of a lane track belt, wherein the direction of the lane track belt refers to the track running direction of the mine car; the point cloud data to be calculated refers to road point data within a preset range, wherein the preset range is set in advance based on the actual road width.
In this embodiment, the purpose of taking the average value of the elevation in the preset neighborhood as the elevation value of the data corresponding to the point is to effectively eliminate the influence of accidental errors, so as to obtain an accurate road elevation change condition, wherein the preset neighborhood is set in advance; the first point cloud data is obtained by replacing the elevation value of the point cloud data to be calculated with the average value of the elevation in the preset neighborhood.
In the embodiment, the preset calculation interval, the preset sampling interval and the preset road slope are set in advance based on the mine road complexity and the actual application requirement and combined with the international flatness index; the gradient value is obtained by calculating dynamic reaction quantity obtained by solving by an iteration method; the first flatness is obtained based on the obtained gradient adjustment values of all the sampling points and the total number of the sampling points in the interval; leveling measurement; the calibration flatness is obtained by using leveling measurement, which is to measure the height difference between two points of the ground by using a leveling instrument and a leveling rod, and can be extracted from a mining database; road flatness is obtained by subtracting a systematic deviation from the first flatness, wherein the systematic deviation refers to the difference between the first flatness and the nominal flatness.
The beneficial effects of the technical scheme are as follows: the accurate road flatness is obtained rapidly by carrying out segmentation calculation on the equidistant neighborhood average value of the road points extracted from the target data by using a cloth simulation algorithm and using the calibrated flatness obtained by leveling calculation, thereby being beneficial to the establishment of a subsequent model.
The embodiment of the invention provides a perception model building method for an open-air mine car, which is used for judging the condition of a road section in front of the open-air mine car in real time by utilizing a road safety perception model to obtain and output a first perception result, and comprises the following steps:
preprocessing point cloud frame data obtained by utilizing a three-dimensional laser radar installed on an outdoor mine car to acquire the condition of a road section in front of the vehicle in real time, and inputting the point cloud frame data into a road safety perception model to obtain a real-time road condition evaluation coefficient of the road in front;
if the real-time road condition evaluation coefficient is not smaller than a preset evaluation threshold, judging and calibrating the safety level of the front road to be three-level, and outputting the safety level as a first perception result;
if the real-time road condition evaluation coefficient is smaller than a preset evaluation threshold value and larger than a preset low threshold value, judging and calibrating the safety level of the front road as a second level, and outputting the second level as a first perception result;
otherwise, calibrating the safety level of the road in front as a first level and outputting the first level as a first sensing result.
In the embodiment, the three-dimensional laser radar is deployed on a car body of a mine car and is used for transmitting detection signals to a target, then the received information reflected from the target is compared with the transmission signals, and related information of the target, such as parameters of the distance, the gesture, the shape and the like of the target, is obtained after proper processing, so that detection, tracking and identification of obstacles on a road are realized; the preprocessing of the point cloud frame data refers to; the road safety perception model is a perception model established by combining road flatness and training a neural network based on target data, and is used for judging real-time road conditions in front of the mine car; the real-time road condition evaluation coefficient is an output result obtained by preprocessing point cloud frame data of the road section condition in front of the vehicle acquired in real time and inputting the processed point cloud frame data into a road safety perception model, and is used for judging the real-time road safety level in front of the vehicle; the first perception result refers to a determination result of the road safety level.
In this embodiment, the preset evaluation threshold and the preset low threshold are set in advance.
In this embodiment, for example, there are mine cars 1 and 2, the corresponding real-time road condition evaluation coefficients are smaller than the preset evaluation threshold, the real-time road condition evaluation coefficient of the mine car 1 is smaller than the preset low threshold, and the real-time road condition evaluation coefficient of the mine car 2 is larger than the preset low threshold, at this time, it is determined that the front road safety level of the mine car 1 is first level, and the front road safety level of the mine car 2 is second level.
The beneficial effects of the technical scheme are as follows: the road condition evaluation coefficient obtained by preprocessing point cloud frame data obtained by collecting the road section conditions in front of the vehicle in real time and inputting the processed point cloud frame data into a road safety perception model is analyzed, so that the front road safety level of the current mine car is effectively determined, a foundation is laid for effective response of the mine car, and the driving safety is ensured.
The embodiment of the invention provides a perception model building method for an open-air mine car, which is used for controlling a mine car to respond according to a first perception result by a management and control center, and comprises the following steps:
the control center station collects and analyzes the first sensing result in real time, and if the safety level of the road in front is three-level, the mine car is controlled to normally run along a preset track according to the normal speed;
if the safety level of the road in front is two-level, calculating to obtain the running state of the corresponding dynamic obstacle based on RTK-GPS data and INS course angle data acquired in real time by the integrated navigation system, and predicting to obtain the corresponding target movement track based on Kalman filtering;
taking the areas except the static obstacle as a feasible area, and analyzing whether the mine car collides with the corresponding dynamic obstacle in the running process of the feasible area by combining the target movement track, and if so, predicting to obtain collision points;
connecting the dynamic barrier with the corresponding collision point position, changing the dynamic barrier into a forbidden area, and selecting a local sub-target node to conduct local obstacle avoidance path planning so as to ensure that the mine car avoids the dynamic barrier until reaching the target node;
if collision does not exist, controlling the mine car to run at a speed reduction in a drivable area when the real-time road flatness of the front road is smaller than a preset flatness threshold value;
if the front road safety level is one level, the control center console controls the mine car to brake emergently.
In the embodiment, the control center is used for controlling the mine car to realize automatic deceleration, detour or emergency braking after the first sensing result is collected and analyzed, so that the driving safety is ensured; the first perception result refers to a road security level obtained by analyzing the result output after the data acquired by the three-dimensional laser radar in real time are input into the road security perception model; the preset track refers to the initial travel route of the mine car.
In this embodiment, the integrated navigation system refers to ins+gps; the purpose of calculating the running state of the corresponding dynamic obstacle by using RTK-GPS data and INS course angle data is to avoid the situation that the running state of the dynamic obstacle is erroneously estimated due to uneven distribution of point clouds in the running process of the dynamic obstacle, wherein the RTK-GPS data refer to satellite positioning data, and the INS course angle data refer to attitude angle data output based on an inertial navigation algorithm; kalman filtering is a highly efficient recursive filter for estimating the state changes of dynamic obstacles; the target motion profile refers to the path of travel of the dynamic obstacle identified in the real-time road ahead of the mine car.
In the embodiment, the collision point refers to a collision position which is obtained by combining the motion trail of the dynamic obstacle obtained by prediction of Kalman filtering and the current running speed of the mine car; the forbidden area is a non-running area obtained by connecting and changing the dynamic barrier and the corresponding collision point of the mine car; the local sub-target nodes are combined with forbidden areas, and the end points of local paths selected from the boundaries of the preset detection range of the mine car are selected; the local obstacle avoidance path planning refers to the realization of dynamic obstacle avoidance while introducing a heuristic function to carry out shortest path planning after predicting collision points; the target node refers to a corresponding end point of the initial driving route of the mine car; the preset flattening threshold is set in advance by combining the actual complexity of the mine road; real-time road flatness refers to the flatness of the real-time road ahead of the mine car.
The beneficial effects of the technical scheme are as follows: the first sensing result is acquired and analyzed through the management and control center desk, the safety level of a real-time road in front of the mine car is obtained, and the mine car is controlled to automatically decelerate, avoid or emergently brake according to the safety level of the road, so that the advancing safety of the mine car is effectively guaranteed.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (8)
1. A perception model building method for an open air mine car vehicle, comprising:
step 1: acquiring historical driving data acquired by utilizing a three-dimensional laser radar, and preprocessing the historical driving data to obtain target data;
step 2: training a neural network based on the target data to obtain a road safety perception model;
step 3: the road safety perception model is utilized to judge the condition of the road section in front of the open-air mine car in real time, and a first perception result is obtained and output;
step 4: and the control center platform controls the mine car to respond according to the first sensing result.
2. The method for constructing a perception model of an open-air mine car according to claim 1, wherein the steps of acquiring the historical traveling data acquired by the three-dimensional laser radar and preprocessing the historical traveling data to obtain the target data include:
step 11: extracting a history point cloud frame acquired by scanning a front road by using a three-dimensional laser radar in the running process of a mine car with a preset amount from a mine database;
step 12: filtering dynamic target point clouds in historical point cloud frames through preprocessing, taking the first two frames in every three historical point cloud frames as historical frames, and taking the rest frames as current frames;
step 13: after the pose change matrixes of the historical frame and the current frame are determined, converting the point cloud data of the historical frame into a point cloud coordinate system of the current frame to obtain a first point cloud frame;
step 14: after the dynamic target point cloud clusters obtained in the preprocessing stage are clustered, the dynamic target point cloud in the first point cloud frame is offset eliminated by utilizing a dynamic target translation method, and a fusion frame is obtained and is output as target data.
3. The method for constructing a perception model of an outdoor mining vehicle according to claim 2, wherein filtering out the historical point cloud frame dynamic target point cloud comprises:
step 21: projecting the point cloud data of the historical point cloud frames into a two-dimensional raster pattern,
step 22: according to the point cloud attribute in the grid, comparing the height difference between each point cloud and the lowest point in the grid with a first threshold value, and dividing the point cloud to obtain a ground point Yun Yufei ground point cloud;
step 23: filtering the ground point cloud, calculating and obtaining the maximum height difference between the rest points in the grids and the first number of the rest points in the grids, and regarding the corresponding grids as barrier grids if the maximum height difference is larger than a preset height threshold value and the first number is larger than a preset number threshold value;
step 24: considering the grids except the obstacle grid in the two-dimensional grid map as a drivable grid;
step 25: and determining a dynamic target point cloud cluster and a static target point cloud cluster from the obstacle point cloud grid according to the cluster characteristics of the obstacle point cloud clusters, and completing identification and filtering of the dynamic target point cloud.
4. The method of claim 1, wherein training the neural network to obtain the road safety perception model based on the target data comprises:
step 31: taking a preset amount of historical driving data as an input party, and constructing a road safety perception neural network;
step 32: performing feature extraction on the target data by using a point cloud target detection algorithm based on deep learning to realize target category identification and endow corresponding candidate frames, so as to obtain a plurality of target frames;
step 33: acquiring the number and size parameters of the corresponding dynamic barriers and static barriers according to the target frame, and determining the barrier identification effect by combining the barrier identification test indexes;
step 34: and training a neural network in road safety perception by using three-quarter target frames as training samples and combining road flatness to obtain a road safety perception model, so as to calculate and obtain a road condition evaluation coefficient, wherein the formula is as follows:
the method comprises the steps of carrying out a first treatment on the surface of the In (1) the->The road condition evaluation coefficient corresponding to the ith target frame is expressed; />Road flatness expressed as a road section corresponding to the ith target frame; />The size of the area denoted as the ith obstacle in the ith target frame; />Denoted as the i-th target intra-frame road surface area; />The preset influence weight of the road flatness on road condition evaluation is expressed; />The preset influence weight of the specific gravity of the area occupied by the barrier-free area on the road surface on the road condition evaluation is expressed; />Expressed as a loss factor generated by environmental impact factors in the calculation of road condition assessment.
5. The method according to claim 4, wherein the step of obtaining the number and size parameters of the corresponding dynamic and static obstacles according to the target frame, and determining the obstacle recognition effect in combination with the obstacle recognition test index comprises the steps of:
randomly selecting a certain amount of target frames as test frames, combining and analyzing the number of dynamic obstacles and static obstacles in the test frames with the number of real obstacles, and obtaining the positive detection rate, the false detection rate and the omission rate of the corresponding obstacles by using a test index formula;
the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>A positive detection rate expressed as an obstacle in the jth target frame; />The correct number of detected obstacles in the jth target frame; />Expressed as the total number of obstacles in the jth target frame; />A false detection rate expressed as an obstacle in the jth target frame; />A false detection number expressed as an obstacle in the jth target frame;a number of obstacles represented as incorrectly classified in the jth target frame; />The false negative rate of the obstacle in the jth target frame; />Expressed as the total number of undetected obstacles in the jth target frame;
and according to the test index result, the false detection rate and the omission factor are both lower than the preset threshold, and the obstacle recognition effect is judged to be good.
6. The method for constructing a perception model for an open air mining vehicle according to claim 4, wherein the process of calculating the road condition evaluation coefficient further comprises: calculating the road flatness, specifically comprising:
step 41: extracting road points from the target data by using a cloth simulation algorithm, creating a track vector line along the track belt direction of the lane, and taking the road points in a preset range as corresponding point cloud data to be calculated;
step 42: acquiring the average value of the elevation in the preset neighborhood of the point cloud data to be calculated as the elevation value of the corresponding data of the point, and acquiring first point cloud data;
step 43: after segmenting the first point cloud data according to a preset calculation interval, acquiring the number of points in each section interval according to a preset sampling interval, so as to calculate and obtain an effective segmentation number;
step 44: the method comprises the steps of adopting a preset road slope as an initial value, solving by an iteration method to obtain 4 dynamic reaction quantities of the speed and the acceleration of the corresponding vehicle body spring bodies and non-spring bodies of the mine car running through all sampling points on the longitudinal and transverse surfaces, and then calculating to obtain an adjustment slope value of each sampling point;
step 45: and according to the gradient adjustment value, calculating to obtain a first flatness, and subtracting a system deviation obtained by comparing the first flatness with the calibrated flatness obtained by leveling calculation to obtain the road flatness.
7. The method for constructing a perception model for an open-air mine car according to claim 1, wherein the real-time judgment of the road conditions of the road section in front of the open-air mine car by using the road safety perception model, the first perception result is obtained and output, comprises:
preprocessing point cloud frame data obtained by utilizing a three-dimensional laser radar installed on an outdoor mine car to acquire the condition of a road section in front of the vehicle in real time, and inputting the point cloud frame data into a road safety perception model to obtain a real-time road condition evaluation coefficient of the road in front;
if the real-time road condition evaluation coefficient is not smaller than a preset evaluation threshold, judging and calibrating the safety level of the front road to be three-level, and outputting the safety level as a first perception result;
if the real-time road condition evaluation coefficient is smaller than a preset evaluation threshold value and larger than a preset low threshold value, judging and calibrating the safety level of the front road as a second level, and outputting the second level as a first perception result;
otherwise, calibrating the safety level of the road in front as a first level and outputting the first level as a first sensing result.
8. A perception model building method for an open air mine car vehicle as claimed in claim 1, wherein the control centre controls the mine car to respond according to the first perception result, comprising:
the control center station collects and analyzes the first sensing result in real time, and if the safety level of the road in front is three-level, the mine car is controlled to normally run along a preset track according to the normal speed;
if the safety level of the road in front is two-level, calculating to obtain the running state of the corresponding dynamic obstacle based on RTK-GPS data and INS course angle data acquired in real time by the integrated navigation system, and predicting to obtain the corresponding target movement track based on Kalman filtering;
taking the areas except the static obstacle as a feasible area, and analyzing whether the mine car collides with the corresponding dynamic obstacle in the running process of the feasible area by combining the target movement track, and if so, predicting to obtain collision points;
connecting the dynamic barrier with the corresponding collision point position, changing the dynamic barrier into a forbidden area, and selecting a local sub-target node to conduct local obstacle avoidance path planning so as to ensure that the mine car avoids the dynamic barrier until reaching the target node;
if collision does not exist, controlling the mine car to run at a speed reduction in a drivable area when the real-time road flatness of the front road is smaller than a preset flatness threshold value;
if the front road safety level is one level, the control center console controls the mine car to brake emergently.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310641511.0A CN116859413A (en) | 2023-05-31 | 2023-05-31 | Perception model building method for open-air mine car |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310641511.0A CN116859413A (en) | 2023-05-31 | 2023-05-31 | Perception model building method for open-air mine car |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116859413A true CN116859413A (en) | 2023-10-10 |
Family
ID=88229367
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310641511.0A Pending CN116859413A (en) | 2023-05-31 | 2023-05-31 | Perception model building method for open-air mine car |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116859413A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117406754A (en) * | 2023-12-01 | 2024-01-16 | 湖北迈睿达供应链股份有限公司 | Logistics robot environment sensing and obstacle avoidance method and system |
| CN117629147A (en) * | 2024-01-25 | 2024-03-01 | 北京易控智驾科技有限公司 | Obstacle detection method, cloud control platform and unmanned vehicle |
-
2023
- 2023-05-31 CN CN202310641511.0A patent/CN116859413A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117406754A (en) * | 2023-12-01 | 2024-01-16 | 湖北迈睿达供应链股份有限公司 | Logistics robot environment sensing and obstacle avoidance method and system |
| CN117406754B (en) * | 2023-12-01 | 2024-02-20 | 湖北迈睿达供应链股份有限公司 | Logistics robot environment sensing and obstacle avoidance method and system |
| CN117629147A (en) * | 2024-01-25 | 2024-03-01 | 北京易控智驾科技有限公司 | Obstacle detection method, cloud control platform and unmanned vehicle |
| CN117629147B (en) * | 2024-01-25 | 2024-03-26 | 北京易控智驾科技有限公司 | Obstacle detection method, cloud control platform and unmanned vehicle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111670468B (en) | Moving body behavior prediction device and moving body behavior prediction method | |
| US10203409B2 (en) | Method and device for the localization of a vehicle from a fixed reference map | |
| US10281920B2 (en) | Planning for unknown objects by an autonomous vehicle | |
| US9127956B2 (en) | Technique for lane assignment in a vehicle | |
| US10234864B2 (en) | Planning for unknown objects by an autonomous vehicle | |
| CN116859413A (en) | Perception model building method for open-air mine car | |
| CN102208013A (en) | Scene matching reference data generation system and position measurement system | |
| US20220197301A1 (en) | Vehicle Localization Based on Radar Detections | |
| Artuñedo et al. | A decision-making architecture for automated driving without detailed prior maps | |
| Lee et al. | Collision risk assessment for possible collision vehicle in occluded area based on precise map | |
| CN113597396A (en) | On-road positioning method and apparatus using road surface characteristics | |
| KR102423550B1 (en) | Method for determining traffic violation vehicle which exists on driving route of autonomous vehicle and the device using the same | |
| JP2021082286A (en) | System and method for improving lane change detection, and non-temporary computer-readable medium | |
| Kellner et al. | Multi-cue, model-based detection and mapping of road curb features using stereo vision | |
| Lee et al. | Ods-bot: Mobile robot navigation for outdoor delivery services | |
| CN109241855A (en) | Intelligent vehicle based on stereoscopic vision can travel area detection method | |
| KR20220073472A (en) | Cross section integrated information providing system and method based on V2X | |
| EP4016129A1 (en) | Radar reference map generation | |
| EP4027169A2 (en) | Radar reference map generation | |
| Worrall et al. | A probabilistic method for detecting impending vehicle interactions | |
| CN114631040A (en) | Apparatus and method for autonomously locating a mobile vehicle on a railway track | |
| Kreibich et al. | Lane-level matching algorithm based on GNSS, IMU and map data | |
| US11651583B2 (en) | Multi-channel object matching | |
| US11592565B2 (en) | Flexible multi-channel fusion perception | |
| CN114078326B (en) | Collision detection method, device, visual sensor and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |