CN115201778B - Irregular obstacle detection method, vehicle and computer-readable storage medium - Google Patents

Irregular obstacle detection method, vehicle and computer-readable storage medium Download PDF

Info

Publication number
CN115201778B
CN115201778B CN202211100133.7A CN202211100133A CN115201778B CN 115201778 B CN115201778 B CN 115201778B CN 202211100133 A CN202211100133 A CN 202211100133A CN 115201778 B CN115201778 B CN 115201778B
Authority
CN
China
Prior art keywords
target
point cloud
cloud data
data
polygon
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.)
Active
Application number
CN202211100133.7A
Other languages
Chinese (zh)
Other versions
CN115201778A (en
Inventor
周萱
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou Xiaopeng Motors Technology Co Ltd
Original Assignee
Guangzhou Xiaopeng Autopilot Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangzhou Xiaopeng Autopilot Technology Co Ltd filed Critical Guangzhou Xiaopeng Autopilot Technology Co Ltd
Priority to CN202211100133.7A priority Critical patent/CN115201778B/en
Publication of CN115201778A publication Critical patent/CN115201778A/en
Application granted granted Critical
Publication of CN115201778B publication Critical patent/CN115201778B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/415—Identification of targets based on measurements of movement associated with the target
    • 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
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
    • G01S13/06—Systems determining position data of a target
    • 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
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
    • G01S13/931—Radar or analogous 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
    • G01S15/00—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems
    • G01S15/02—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems using reflection of acoustic waves
    • G01S15/06—Systems determining the position data of a target
    • 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
    • G01S15/00—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems
    • G01S15/88—Sonar systems specially adapted for specific applications
    • G01S15/93—Sonar systems specially adapted for specific applications for anti-collision purposes
    • G01S15/931—Sonar 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
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves
    • G01S17/06—Systems determining position data of a target
    • 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
    • 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
    • G01S7/52—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S15/00
    • G01S7/539—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S15/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • 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
    • 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
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/165—Anti-collision systems for passive traffic, e.g. including static obstacles, trees
    • 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)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Theoretical Computer Science (AREA)
  • Electromagnetism (AREA)
  • Multimedia (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Medical Informatics (AREA)
  • Artificial Intelligence (AREA)
  • Acoustics & Sound (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Human Computer Interaction (AREA)
  • Traffic Control Systems (AREA)
  • Optical Radar Systems And Details Thereof (AREA)

Abstract

The invention provides an irregular obstacle detection method for a vehicle, the vehicle and a computer-readable storage medium. The obstacle detection method includes: screening to obtain a plurality of target cubes and a plurality of target polygons according to the target point cloud data; wherein the target cube and the target polygon contain spatial position information of the obstacle; detecting a target polygon in a detection boundary value preset by a target cube to generate a target boundary; and obtaining obstacle range data according to the target boundary so as to detect irregular obstacles. The obstacle detection method provided by the invention can detect obstacles according to all parts corresponding to the target point cloud data, and calibrate the detected obstacles by judging the target boundary corresponding to the combination of the target cube and the target polygon.

Description

Irregular obstacle detection method, vehicle, and computer-readable storage medium
Technical Field
The present invention relates to the field of vehicle technologies, and more particularly, to an irregular obstacle detection method for a vehicle, and a computer-readable storage medium.
Background
The related art can detect regular obstacles on a road during the driving of a vehicle, but the detection method for detecting obstacles with irregular shapes, such as vehicles carrying irregular-shaped cargos, pedestrians carrying cargos and the like, is less involved. Therefore, how to accurately detect irregular obstacles on the road in the driving process of the vehicle becomes a problem to be solved urgently.
Disclosure of Invention
The invention provides an irregular obstacle detection method for a vehicle, the vehicle and a computer-readable storage medium.
The invention provides an irregular obstacle detection method for a vehicle, which comprises the following steps:
screening according to the target point cloud data to obtain a plurality of target cubes and a plurality of target polygons; wherein the target cube and the target polygon contain spatial location information of an obstacle;
detecting the target polygon in a detection boundary value preset by the target cube to generate a target boundary;
and obtaining obstacle range data according to the target boundary so as to detect the irregular obstacle.
Therefore, the method firstly screens and obtains the recognizable regular obstacles and irregular obstacles according to the target point cloud data, and generates a target cube to fit the recognizable regular obstacles and a target polygon to fit the irregular obstacles. And further, detecting whether a target polygon exists in the detection boundary value range of the target cube by taking the target cube as a reference, so as to judge whether the target cube and the target polygon need to be combined together, and further generating a target boundary according to a judgment result. And then obstacle range data corresponding to the obstacle can be generated according to the target boundary, so that the position and the size of the obstacle are calibrated according to the obstacle range data, and the obstacle is detected. The obstacle detection method provided by the invention can detect the obstacles near the vehicle, particularly the space position of random obstacles appearing on the lane and the space range occupied by the obstacles, so as to timely detect the random obstacles appearing on the lane which dynamically changes in real time, and improve the driving safety in the driving process of the vehicle.
In some embodiments, the obtaining a plurality of target cubes and a plurality of target polygons by filtering according to the target point cloud data includes:
acquiring initial point cloud data, and preprocessing the initial point cloud data according to target positioning information to obtain target point cloud data;
and optimizing and screening the target point cloud data through a preset deep learning model to obtain a plurality of target cubes and a plurality of target polygons.
Therefore, the initial point cloud data is processed through the target positioning information to screen out noise points in the target point cloud data or improve the effectiveness of part of point clouds in the target point cloud data, and then the target point cloud data is obtained. Further, the method comprises the steps of optimizing target point cloud data again through a preset deep learning model, and carrying out reasoning calculation on the target point cloud data according to the deep learning model to distinguish point clouds of regular obstacle parts and point clouds of irregular obstacle parts, so that a target cube capable of fitting the regular obstacle point clouds and a target polygon capable of fitting the irregular obstacle point clouds are generated.
In some embodiments, the screening the target point cloud data through a preset deep learning model optimization to obtain a plurality of target cubes and a plurality of target polygons includes:
optimizing the target point cloud data according to the deep learning model to obtain first point cloud data and second point cloud data;
generating a plurality of target cubes according to the first point cloud data;
and processing the second point cloud data through a preset clustering algorithm to obtain a plurality of target polygons.
Therefore, the target point cloud data are classified through the deep learning model, and first point cloud data containing a target cube fitting object and second point cloud data of the target point cloud data except for the first point cloud data are obtained. Further, the method obtains the shape of the first point cloud data, and selects a proper target cube to fit the first point cloud data, so that the target cube can completely and exactly contain all the first point cloud data. Further, the method carries out clustering analysis on the second point cloud data through a clustering algorithm to screen out the point cloud data of the accessory part of the suspected obstacle, and selects a proper target polygon to fit the part of point cloud data by adopting a selection method similar to that of a target cube. Thus, a target cube and a target polygon are obtained.
In some embodiments, the generating a number of the target cubes from the first point cloud data comprises:
optimizing the target point cloud data according to the deep learning model to obtain the first point cloud data and the point cloud orientation data;
generating all cubes according to the first point cloud data;
and screening all cubes according to the vehicle position and the point cloud orientation data to obtain a plurality of target cubes.
In this way, the target point cloud data is optimized through the deep learning model to obtain the first point cloud data and the point cloud orientation data contained in the first point cloud data. The point cloud orientation data can represent orientation information of all cubes in the first point cloud data. Through the point cloud orientation data, the method can screen out the target cube facing to the position of the vehicle from all cubes generated according to the first point cloud data, and then effectively screen out the obstacle.
In some embodiments, the processing the second point cloud data by a preset clustering algorithm to obtain a plurality of target polygons includes:
processing the second point cloud data through the clustering algorithm to obtain all polygons;
and screening all polygons in the range formed by the target cubes by taking the position of the vehicle as the center to obtain a plurality of target polygons.
And further, in an area range formed by surrounding all target cubes by taking the position of the vehicle as the center, selecting target polygons in all the polygons so as to effectively reduce the number of the polygons and improve the detection efficiency of the method.
In some embodiments, detecting the target polygon within a detection boundary value preset by the target cube to generate a target boundary includes:
detecting the target polygon within the preset detection boundary value range; wherein the detection boundary values are matched with the target polygons one by one;
under the condition that the target polygon exists in the detection boundary value range, detecting whether the target polygon exists on a detection boundary corresponding to the detection boundary value;
under the condition that the target polygon exists on the detection boundary, increasing the detection boundary value until the target polygon does not exist on the detection boundary, and generating a target boundary value according to the increased detection boundary value;
and generating a target boundary according to the target boundary value.
In this way, whether a target polygon exists inside the detection boundary is determined according to the detection boundary value, and in the case that the target polygon exists inside the detection boundary, whether the target polygon exists on the detection boundary is further checked. When the target polygon exists on the detection boundary, the method can judge that the current detection boundary can not contain the combination of the target polygon and the target cube, so that the detection boundary value is increased, whether the target polygon is detected on the boundary or not is detected by the larger detection boundary, the target boundary value which can contain the combination of the target polygon and the target cube is obtained, and the target boundary is defined according to the target boundary value.
In some embodiments, the detecting whether the target polygon exists on the detection boundary corresponding to the detection boundary value in the case that the target polygon exists within the detection boundary value range includes:
ground screening is carried out on the point cloud of the target polygon to obtain third point cloud data;
carrying out direction division on the third point cloud according to a preset direction to obtain fourth point cloud data;
and acquiring picture data, and projecting the fourth point cloud data to the picture data to detect whether the target polygon exists on a detection boundary corresponding to the detection boundary value in the preset direction.
Therefore, the point cloud correspondingly fitted to the target polygon is ground screened out to reduce data errors of the point cloud, and third point cloud data with higher precision and effectiveness are obtained.
In some embodiments, the obtaining the picture data and projecting the fourth point cloud data to the picture data to detect whether the target polygon exists on a detection boundary corresponding to the detection boundary value in the preset direction includes:
projecting the fourth point cloud data to the picture data to obtain fourth point cloud plane data;
and performing matching detection according to the fourth point cloud plane data, and detecting whether the target polygon exists on the detection boundary in the preset direction or not under the condition that the matching is passed.
In this way, the fourth point cloud data is projected to the picture data to obtain fourth point cloud plane data, wherein the fourth point cloud plane data simultaneously contains obstacle position information corresponding to the fourth point cloud data and obstacle position information contained in the image data. Further, according to the fourth point cloud plane data, whether the obstacle position information in the image data and the fourth point cloud data is matched or not can be matched through the method, when the fourth point cloud data contains the position of the obstacle information and the obstacle information is also reflected in the image data, the situation that the matching is passed is achieved, and whether a target polygon exists on the detection boundary in each direction or not is detected.
The invention also provides a vehicle comprising a processor and a memory, the memory having stored thereon a computer program which, when executed by the processor, implements the obstacle detection method of any one of the above.
The invention also provides a computer-readable storage medium, wherein computer-executable instructions are stored in the computer-readable storage medium, and a computer can execute the computer-executable instructions to realize the obstacle detection method.
Additional aspects and advantages of embodiments of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of embodiments of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a schematic flow diagram of the obstacle detection method of the present invention;
FIG. 2 is a schematic view of an obstacle of the present invention;
FIG. 3 is a detailed schematic diagram of the obstacle detection method of the present invention;
fig. 4 is a schematic structural view of the vehicle of the invention.
Description of the drawings: vehicle 10, processor 11, memory 12.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary only for explaining the embodiments of the present invention, and are not construed as limiting the embodiments of the present invention.
Referring to fig. 1, the present invention provides an irregular obstacle detection method for a vehicle, including:
s10, screening according to the target point cloud data to obtain a plurality of target cubes and a plurality of target polygons. Wherein the target cube and the target polygon contain spatial position information of the obstacle.
And S20, detecting the target polygon in the detection boundary value preset by the target cube to generate a target boundary.
And S30, obtaining obstacle range data according to the target boundary so as to detect irregular obstacles.
The method comprises the steps of screening according to target point cloud data to obtain identifiable regular obstacles and irregular obstacles, generating a target cube to fit the identifiable regular obstacles, and generating a target polygon to fit the irregular obstacles. And further, detecting whether a target polygon exists in the detection boundary value range of the target cube by taking the target cube as a reference, thereby judging whether the target cube and the target polygon need to be combined together, and further generating a target boundary according to a judgment result. Furthermore, obstacle range data corresponding to the obstacle can be generated according to the target boundary, so that the position and the size of the obstacle are calibrated according to the obstacle range data, and the obstacle is detected. The obstacle detection method can detect obstacles near the vehicle, such as an open vehicle door shown in fig. 2, a loaded electric vehicle, luggage in the hands of a pedestrian, objects held by the pedestrian and the like, particularly the spatial position of random obstacles appearing on a lane and the spatial range occupied by the obstacles, so as to timely detect the random obstacles appearing on the lane which dynamically changes in real time, and improve the driving safety of the vehicle in the driving process.
The vehicle can obtain a target cube and a target polygon through screening according to the target point cloud data. The target cube and the target polygon contain space position information of the obstacle, and the target polygon in a detection boundary value preset by the target cube is detected to generate a target boundary, so that obstacle range data are further obtained according to the target boundary to detect the irregular obstacle.
Specifically, in the driving process of the vehicle, objects existing on the current road can be detected through a laser radar, a millimeter wave radar, an ultrasonic radar and the like on the vehicle, and the detection result is screened to obtain target point cloud information containing all obstacles, so that the obstacles which can be clearly detected are screened out according to the target point cloud data, for example, a vehicle body, a pedestrian, an electric vehicle and the like of the vehicle in fig. 2 are taken as fitting objects of a target cube, and further, extension parts of the suspected target cube are screened out, for example, a vehicle door opened by the vehicle, luggage of the pedestrian, goods loaded on the electric vehicle and the like are taken as the target cube. Therefore, the method can detect the target polygon in the detection boundary value by taking the target cube as a reference so as to judge the connection relation between the target cube and the target polygon, thereby obtaining the target boundary of a set consisting of the target cube and the target polygon, and further converting the target boundary into the obstacle range data so as to calibrate the detected obstacle.
Specifically, the method can output the obstacle range data to the downstream, the downstream can be a related control platform for automatic driving, and the downstream can perform corresponding path planning on the automatic driving of the vehicle according to the obstacle range data provided by the method.
In certain embodiments, step S10 comprises:
and acquiring initial point cloud data, and preprocessing the initial point cloud data according to the target positioning information to obtain target point cloud data.
And optimizing and screening target point cloud data through a preset deep learning model to obtain a plurality of target cubes and a plurality of target polygons.
Therefore, the initial point cloud data can be processed through the target positioning information to screen out noise points in the target point cloud data or improve the effectiveness of part of point clouds in the target point cloud data, and then the target point cloud data is obtained. Further, the target point cloud data is optimized again through a preset deep learning model, and inference calculation is carried out on the target point cloud data according to the deep learning model so as to distinguish the point cloud of the regular obstacle part and the point cloud of the irregular obstacle part, and therefore a target cube capable of fitting the regular obstacle point cloud and a target polygon capable of fitting the irregular obstacle point cloud are generated.
The vehicle can acquire initial point cloud data through the processor, and preprocess the initial point cloud data according to the target positioning information to obtain target point cloud data. And the target point cloud data can be optimized and screened through a preset deep learning model to obtain a plurality of target cubes and a plurality of target polygons.
Specifically, the method can perform motion compensation on the initial point cloud data at different times by using the target positioning information, or perform operations such as down-sampling on the initial point cloud data, so as to achieve the effects of improving the effectiveness of the initial point cloud data and generating the target point cloud data. By the method, the effectiveness of the point cloud data can be effectively improved, the target cube and the target polygon are effectively screened and generated through the deep learning model, unnecessary calculation processes are reduced, and the detection efficiency of the obstacle is improved.
In some embodiments, the method for optimizing and screening target point cloud data through a preset deep learning model to obtain a plurality of target cubes and a plurality of target polygons includes:
and optimizing the target point cloud data according to the deep learning model to obtain first point cloud data and second point cloud data.
And generating a plurality of target cubes according to the first point cloud data.
And processing the second point cloud data through a preset clustering algorithm to obtain a plurality of target polygons.
Therefore, the target point cloud data can be classified through the deep learning model, and first point cloud data containing a target cube fitting object and second point cloud data except for the first point cloud data part in the target point cloud data are obtained. Further, the shape of the first point cloud data is obtained, and a suitable target cube is selected to fit the first point cloud data, so that the target cube can completely and exactly contain all the first point cloud data. Further, the second point cloud data is subjected to clustering analysis through a clustering algorithm to screen out the point cloud data of the accessory part of the suspected obstacle, and a suitable target polygon is selected to fit the part of the point cloud data by adopting a selection method similar to that of a target cube. Thus, a target cube and a target polygon are obtained.
The vehicle can optimize the target point cloud data according to the deep learning model to obtain first point cloud data and second point cloud data. Several target cubes may also be generated from the first point cloud data. And processing the second point cloud data through a preset clustering algorithm to obtain a plurality of target polygons.
By the method, the point cloud data containing the characteristics of all parts in the target point cloud data can be effectively distinguished to obtain the first point cloud data and the second point cloud data, then the obstacle main body on the road is fitted through the first point cloud data to obtain the target cube, and further the point cloud data connected with the obstacle main body part in the second point cloud data is screened to obtain the target polygon through fitting, so that the target cube and the target polygon are provided.
In some embodiments, generating a number of target cubes from the first point cloud data includes:
and optimizing the target point cloud data according to the deep learning model to obtain first point cloud data and point cloud orientation data.
And generating all cubes according to the first point cloud data.
And screening all cubes according to the vehicle position and the point cloud orientation data to obtain a plurality of target cubes.
In this way, the target point cloud data can be optimized through the deep learning model to obtain the first point cloud data and the point cloud orientation data contained in the first point cloud data. Wherein the point cloud orientation data may characterize orientation information of all cubes in the first point cloud data. Through the point cloud orientation data, the method can screen out the target cube facing to the position of the vehicle from all cubes generated according to the first point cloud data, and further screen out the obstacles effectively.
The vehicle can optimize the target point cloud data according to the deep learning model to obtain first point cloud data and point cloud orientation data. All cubes may also be generated from the first point cloud data. And screening all cubes according to the vehicle position and the point cloud orientation data to obtain a plurality of target cubes.
According to the invention, all cubes obtained by screening can be screened through the point cloud orientation data while the target point cloud data processed by the deep learning model is utilized, so that the target cube which causes obstacles to the driving process of the vehicle is obtained. Further, obstacle detection is performed on all cubes, so that the detection efficiency of the obstacle detection method provided by the invention is influenced.
In some embodiments, the second point cloud data is processed by a preset clustering algorithm to obtain a plurality of target polygons, including:
and processing the second point cloud data through a clustering algorithm to obtain all polygons.
And screening all polygons in the range formed by the target cubes by taking the position of the vehicle as the center to obtain a plurality of target polygons.
Therefore, the second point cloud data can be subjected to clustering analysis through a clustering algorithm to obtain point clouds corresponding to irregular obstacle parts in the second point cloud data, all polygons are obtained by adopting the polygons to fit the point clouds one by one, and further, target polygons in all polygons are selected in an area range formed by all surrounding target cubes by taking the position of the vehicle as the center, so that the number of the polygons is effectively reduced, and the detection efficiency of the method is improved.
The vehicle can process the second point cloud data through a clustering algorithm to obtain all polygons. And screening all polygons in the range formed by the target cubes by taking the position of the vehicle as the center to obtain a plurality of target polygons.
Specifically, referring to fig. 3, the method may select, as the target polygon, a portion (such as the dashed frames 1, 2, 3, and 4) of the total polygons within a range area formed by the surrounding target cubes according to the position of the vehicle as the detection center, so as to screen out a target polygon that meets the requirement (i.e., screen out the dashed frame 5, and reserve the dashed frames 1, 2, 3, and 4), thereby improving the detection efficiency of the method.
In certain embodiments, S20 comprises:
detecting a target polygon within a detection boundary value range, wherein the detection boundary values correspond to the target polygons in a one-to-one mode;
and under the condition that the target polygon exists in the detection boundary value range, detecting whether the target polygon exists on a detection boundary corresponding to the detection boundary value.
And under the condition that the target polygon exists on the detection boundary, increasing the detection boundary value until the target polygon does not exist on the detection boundary, and generating a target boundary value according to the increased detection boundary value.
And generating a target boundary according to the target boundary value.
In this way, whether a target polygon exists inside the detection boundary can be detected according to the detection boundary value, and in the case that the target polygon exists inside the detection boundary, whether the target polygon exists on the detection boundary can be further checked. When a target polygon exists on the detection boundary, it can be determined that the current detection boundary cannot contain the combination of the target polygon and the target cube, so as to increase the detection boundary value, and detect whether the target polygon is on the boundary with a larger detection boundary, so as to obtain a target boundary value which can contain the combination of the target polygon and the target cube, and further define the target boundary according to the target boundary value. In addition, the detection boundary values provided by the method are matched with the target polygons one by one, and each target polygon corresponds to one detection boundary value, so that the method can extract the corresponding detection boundary values according to the characteristics of each target polygon, and can more accurately fit the target boundaries.
The vehicle may detect the target polygon within a range of detection boundary values, where the detection boundary values correspond to the target polygons on a one-to-one basis. And under the condition that the target polygon exists in the boundary value range, detecting whether the target polygon exists on a detection boundary corresponding to the detection boundary value. And under the condition that the target polygon exists on the detection boundary, increasing the detection boundary value until the target polygon does not exist on the detection boundary, generating a target boundary value according to the increased detection boundary value, and generating a target boundary according to the target boundary value. The minimum detection boundary value may be a close proximity to the object cube, and the maximum detection boundary value may be a predetermined range, for example, 1 meter, larger than the object cube boundary. In the case where a plurality of target polygons exist within the boundary value range, it is further detected whether a target polygon exists on the boundary of each polygon.
By the method, the target boundary value which can contain the combination of the target polygon and the target cube can be obtained through gradual fitting, the size and the spatial position of the detected obstacle are defined through the target boundary value, and the judgment and the detection of the obstacle near the vehicle are realized.
In some embodiments, in the case that a target polygon exists in the boundary value range, detecting whether the target polygon exists on a detection boundary corresponding to the detection boundary value includes:
and (4) carrying out ground screening on the point cloud of the target polygon to obtain third point cloud data.
And carrying out direction division on the third point cloud according to a preset direction to obtain fourth point cloud data.
And acquiring picture data, and projecting the fourth point cloud data to the picture data to detect whether a target polygon exists on a detection boundary corresponding to the boundary value in a preset direction.
Therefore, the point clouds correspondingly fitted to the target polygons can be ground screened out, so that data errors of the point clouds are reduced, third point cloud data with higher precision and higher effectiveness are obtained, further, the method divides the third point clouds according to the preset direction to obtain fourth point cloud data, and whether the target polygons exist in all directions is judged according to the projection result of the fourth point cloud data projected to the picture annual data to detect whether obstacles exist in the preset direction.
The vehicle can carry out ground screening to the point cloud of target polygon, obtains third point cloud data. And direction division can be carried out on the third point cloud according to a preset direction to obtain fourth point cloud data. And further, image data can be obtained, and the fourth point cloud data is projected to the image data so as to detect whether a target polygon exists on a detection boundary corresponding to the boundary value in the preset direction.
Specifically, the method can perform ground real-position processing on the point cloud fitted by the target point cloud data to obtain third point cloud data, split the third point cloud data into fourth point cloud data in four directions according to the preset four directions of front, back, left and right, and project the fourth point cloud data in each direction to the picture data in the corresponding direction to obtain two-dimensional images in the four directions, so as to realize detection of the target polygon in the detection boundary value range. The detection of obstacles near the vehicle is effectively realized.
Specifically, since the proportion of part of point cloud data is small, the corresponding polygon cannot be obtained through fitting, so that the method can project the initial point cloud data in each direction to the image data corresponding to each direction, the probability of the method for detecting the obstacle failure is reduced, and the detection accuracy of the method is ensured.
In some embodiments, acquiring the picture data and projecting the fourth point cloud data to the picture data to detect whether a target polygon exists on a detection boundary corresponding to the boundary value in a preset direction includes:
and projecting the fourth point cloud data to the picture data to obtain fourth point cloud plane data.
And performing matching detection according to the fourth point cloud plane data, and detecting whether a target polygon exists on the boundary in a preset direction under the condition that the matching is passed.
Thus, the fourth point cloud data can be projected to the picture data to obtain fourth point cloud plane data, where the fourth point cloud plane data includes both obstacle position information corresponding to the fourth point cloud data and obstacle position information included in the image data. Further, according to the fourth point cloud plane data, whether the obstacle position information in the image data and the fourth point cloud data is matched or not can be matched through the method, when the fourth point cloud data contains the position of the obstacle information and the obstacle information is also reflected in the image data, the situation that the matching is passed is achieved, and whether a target polygon exists on the detection boundary in each direction or not is detected.
The vehicle may project the fourth point cloud data to the picture data to obtain fourth point cloud plane data. And matching detection can be carried out according to the fourth point cloud plane data, and whether a target polygon exists on the boundary or not is detected in the preset direction under the condition that matching is passed.
Specifically, the method obtains fourth point cloud plane data after projecting the fourth point cloud data in each direction to the image data corresponding to each direction. The method further defines a detection boundary on the fourth cloud plane data according to the detection boundary value, and detects whether the projection of the point cloud data corresponding to the target polygon on the detection boundary exists on the image data, so as to detect whether the target polygon exists on the detection boundary.
Referring to fig. 3, in some specific embodiments, the method may implement obstacle recognition on a road based on a radar, a positioning device, and a camera on the host vehicle. Firstly, the method can acquire the point cloud data at the moment through the radar drive, and performs data optimization on the point cloud data through the positioning information output by the positioning equipment. Meanwhile, the camera can acquire the current road condition and analyze the road condition to obtain a range 2D frame, a predicted 3D position and an obstacle orientation of the obstacle on the road.
Further, the method comprises the steps of carrying out data processing on the point cloud data after data optimization through a deep learning model to obtain a target cube used for representing the size and position information of the obstacle body, obtaining the point cloud possibly connected with the obstacle through a clustering algorithm again, and generating a target polygon to fit the part of the point cloud.
Further, the method projects the point cloud data to image data acquired by a camera, generates a two-dimensional image simultaneously containing obstacle position information detected by the radar and the camera, and can judge that the obstacle information exists and execute further detection when the obstacle position information detected by the two parts is overlapped.
Further, the method detects a possible connection relationship between the acquired target polygon and the target cube, determines that the target polygon and the target cube belong to the same obstacle when the distance between the target polygon and the target cube is within a preset range, such as a combination of a dashed line frame 40002 and a dashed line frame 2, a dashed line frame 3 in fig. 3, or a combination of a dashed line frame 40003 and a dashed line frame 4, and generates a target boundary to completely frame and select the combination of the target cube and the target polygon to represent the spatial position and size of the detected obstacle, such as a solid line frame including the combination of the dashed line frame 40002 and the dashed line frame 2, the dashed line frame 3 in fig. 3, and a combination including the dashed line frame 40003 and the dashed line frame 4. Or when the distance between the target polygon and the target cube is greater than the preset range, the method determines that the target cube and the target polygon do not have a connection relationship and do not belong to the same obstacle, such as the dashed line frame 40001 and the dashed line frame 1.
The invention also provides a vehicle 10, the vehicle 10 comprising a processor 11 and a memory 12, the memory 12 storing a computer program, the computer program when executed by the processor 11 implementing any of the above obstacle detection methods.
The vehicle 10 provided by the present invention may execute 11 instructions stored in memory 12 to implement an obstacle detection method as provided by the present invention. The beneficial effects of the obstacle detection method provided by the invention have been described in detail in the foregoing, and are not described herein again.
The invention also provides a computer-readable storage medium, wherein computer-executable instructions are stored in the computer-readable storage medium, and a computer can execute the computer-executable instructions to realize the obstacle detection method of any one of the above items.
The present invention provides a computer-readable storage medium in which a computer program is stored, and by executing this computer program, a computer can implement the obstacle detection method as provided by the present invention. The beneficial effects of the obstacle detection method provided by the invention are elaborated in the foregoing, and are not described in detail herein.
The invention also provides a non-transitory computer-readable storage medium containing the computer program. The computer program, when executed by one or more processors, causes the one or more processors to perform the interaction method of the present invention.
In the description herein, references to the description of "certain embodiments," "one example," "exemplary," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps in the process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and not to be construed as limiting the present invention, and those skilled in the art can make changes, modifications, substitutions and alterations to the above embodiments within the scope of the present invention.

Claims (10)

1. An irregular obstacle detection method for a vehicle, characterized by comprising:
screening to obtain a plurality of target cubes and a plurality of target polygons according to the target point cloud data; wherein the target cube and the target polygon contain spatial position information of an obstacle, the target cube fits recognizable regular obstacles, and the target polygon fits irregular obstacles;
detecting whether the target polygon exists in a detection boundary value range preset by the target cube, and judging whether the target cube and the target polygon need to be combined together so as to generate a target boundary of a set formed by the target cube and the target polygon;
and obtaining obstacle range data according to the target boundary so as to detect the irregular obstacle.
2. The irregular obstacle detection method according to claim 1, wherein the screening a plurality of target cubes and a plurality of target polygons according to the target point cloud data comprises:
acquiring initial point cloud data, and preprocessing the initial point cloud data according to target positioning information to obtain target point cloud data;
and optimizing and screening the target point cloud data through a preset deep learning model to obtain a plurality of target cubes and a plurality of target polygons.
3. The irregular obstacle detection method according to claim 2, wherein the step of optimizing and screening the target point cloud data through a preset deep learning model to obtain a plurality of target cubes and a plurality of target polygons comprises:
optimizing the target point cloud data according to the deep learning model to obtain first point cloud data and second point cloud data;
generating a plurality of target cubes according to the first point cloud data;
and processing the second point cloud data through a preset clustering algorithm to obtain a plurality of target polygons.
4. The irregular obstacle detection method of claim 3, wherein the generating a number of the target cubes from the first point cloud data comprises:
optimizing the target point cloud data according to the deep learning model to obtain the first point cloud data and point cloud orientation data;
generating all cubes according to the first point cloud data;
and screening all cubes according to the vehicle position and the point cloud orientation data to obtain a plurality of target cubes.
5. The irregular obstacle detection method according to claim 3, wherein the processing the second point cloud data by a predetermined clustering algorithm to obtain a plurality of target polygons comprises:
processing the second point cloud data through the clustering algorithm to obtain all polygons;
and screening all polygons in the range formed by the target cubes by taking the position of the vehicle as the center to obtain a plurality of target polygons.
6. The irregular obstacle detection method according to claim 1, wherein the detecting the target polygon within a detection boundary value preset by the target cube to generate a target boundary comprises:
detecting the target polygon within the preset detection boundary value range; wherein the detection boundary values are matched with the target polygons one by one;
under the condition that the target polygon exists in the detection boundary value range, detecting whether the target polygon exists on a detection boundary corresponding to the detection boundary value or not;
under the condition that the target polygon exists on the detection boundary, increasing the detection boundary value until the target polygon does not exist on the detection boundary, and generating a target boundary value according to the increased detection boundary value;
and generating a target boundary according to the target boundary value.
7. The irregular obstacle detection method according to claim 6, wherein the detecting whether the target polygon exists on the detection boundary corresponding to the detection boundary value in the case where the target polygon exists within the detection boundary value range includes:
ground screening is carried out on the point cloud of the target polygon to obtain third point cloud data;
carrying out direction division on the third point cloud according to a preset direction to obtain fourth point cloud data;
and acquiring picture data, and projecting the fourth point cloud data to the picture data to detect whether the target polygon exists on a detection boundary corresponding to the detection boundary value in the preset direction.
8. The irregular obstacle detection method according to claim 7, wherein the acquiring picture data and projecting the fourth point cloud data onto the picture data to detect whether the target polygon exists on a detection boundary corresponding to the detection boundary value in the preset direction comprises:
projecting the fourth point cloud data to the picture data to obtain fourth point cloud plane data;
and performing matching detection according to the fourth point cloud plane data, and detecting whether the target polygon exists on the detection boundary in the preset direction or not under the condition that the matching is passed.
9. A vehicle, characterized in that the vehicle comprises a processor and a memory, in which a computer program is stored which, when being executed by the processor, carries out the obstacle detection method according to any one of claims 1-8.
10. A computer-readable storage medium having computer-executable instructions stored thereon, the computer-executable instructions being executable by a computer to implement the obstacle detection method of any one of claims 1 to 8.
CN202211100133.7A 2022-09-09 2022-09-09 Irregular obstacle detection method, vehicle and computer-readable storage medium Active CN115201778B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211100133.7A CN115201778B (en) 2022-09-09 2022-09-09 Irregular obstacle detection method, vehicle and computer-readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211100133.7A CN115201778B (en) 2022-09-09 2022-09-09 Irregular obstacle detection method, vehicle and computer-readable storage medium

Publications (2)

Publication Number Publication Date
CN115201778A CN115201778A (en) 2022-10-18
CN115201778B true CN115201778B (en) 2023-02-03

Family

ID=83572841

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211100133.7A Active CN115201778B (en) 2022-09-09 2022-09-09 Irregular obstacle detection method, vehicle and computer-readable storage medium

Country Status (1)

Country Link
CN (1) CN115201778B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118135543A (en) * 2024-05-06 2024-06-04 成都赛力斯科技有限公司 Parking space and obstacle detection method and device, electronic equipment and storage medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110095062A (en) * 2019-04-17 2019-08-06 北京华捷艾米科技有限公司 A kind of object volume measurement method of parameters, device and equipment
CN112489123A (en) * 2020-10-30 2021-03-12 江阴市智行工控科技有限公司 Three-dimensional positioning method for surface target of truck in steel mill reservoir area
CN113511503A (en) * 2021-06-17 2021-10-19 北京控制工程研究所 Independent intelligent method for collecting, collecting and collecting uncertain objects in extraterrestrial exploration
CN113610916A (en) * 2021-06-17 2021-11-05 同济大学 Irregular object volume determination method and system based on point cloud data
CN113959331A (en) * 2021-06-03 2022-01-21 武汉万集信息技术有限公司 Object packaging volume measuring method and device
CN114488181A (en) * 2022-01-07 2022-05-13 北京理工大学 Multi-source heterogeneous sensor fusion method and device for camera and laser radar
CN114882198A (en) * 2022-06-08 2022-08-09 一汽解放汽车有限公司 Target determination method, device, equipment and medium

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10646999B2 (en) * 2017-07-20 2020-05-12 Tata Consultancy Services Limited Systems and methods for detecting grasp poses for handling target objects
US20210223363A1 (en) * 2020-01-16 2021-07-22 Outsight SA Object detection on a path of travel and obstacle detection on railway tracks using free space information
CN111674390B (en) * 2020-05-09 2021-07-02 惠州市德赛西威汽车电子股份有限公司 Obstacle avoidance method for automatic parking path planning and parking path planning system
CN112356019B (en) * 2020-08-06 2024-06-14 武汉科技大学 Object shape analysis method and device for smart hand grabbing

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110095062A (en) * 2019-04-17 2019-08-06 北京华捷艾米科技有限公司 A kind of object volume measurement method of parameters, device and equipment
CN112489123A (en) * 2020-10-30 2021-03-12 江阴市智行工控科技有限公司 Three-dimensional positioning method for surface target of truck in steel mill reservoir area
CN113959331A (en) * 2021-06-03 2022-01-21 武汉万集信息技术有限公司 Object packaging volume measuring method and device
CN113511503A (en) * 2021-06-17 2021-10-19 北京控制工程研究所 Independent intelligent method for collecting, collecting and collecting uncertain objects in extraterrestrial exploration
CN113610916A (en) * 2021-06-17 2021-11-05 同济大学 Irregular object volume determination method and system based on point cloud data
CN114488181A (en) * 2022-01-07 2022-05-13 北京理工大学 Multi-source heterogeneous sensor fusion method and device for camera and laser radar
CN114882198A (en) * 2022-06-08 2022-08-09 一汽解放汽车有限公司 Target determination method, device, equipment and medium

Also Published As

Publication number Publication date
CN115201778A (en) 2022-10-18

Similar Documents

Publication Publication Date Title
US10366310B2 (en) Enhanced camera object detection for automated vehicles
US10832064B2 (en) Vacant parking space detection apparatus and vacant parking space detection method
KR102108953B1 (en) Robust camera and lidar sensor fusion method and system
Ferryman et al. Visual surveillance for moving vehicles
CN110674705B (en) Small-sized obstacle detection method and device based on multi-line laser radar
KR102269750B1 (en) Method for Real-time Object Detection Based on Lidar Sensor and Camera Using CNN
KR20200022001A (en) Rare Instance Classifiers
WO2020132082A1 (en) Object classification using extra-regional context
CN115201778B (en) Irregular obstacle detection method, vehicle and computer-readable storage medium
EP3555854B1 (en) A method of tracking objects in a scene
CN114384492B (en) Point cloud processing method and device for laser radar and storage medium
US20220171975A1 (en) Method for Determining a Semantic Free Space
JP7418476B2 (en) Method and apparatus for determining operable area information
KR20170106823A (en) Image processing device identifying object of interest based on partial depth map
Steinemann et al. Geometric-model-free tracking of extended targets using 3D lidar measurements
EP4099211A1 (en) Method and device for training a machine learning algorithm
CN114140660A (en) Vehicle detection method, device, equipment and medium
US20240193803A1 (en) Apparatus and method for object detection
KR102672992B1 (en) Method for determining whether camera is contaminated in real time and device using the same
US20230351741A1 (en) Method and device for training a neural network
US20230008457A1 (en) Occupancy Grid Calibration
Ozutemiz et al. Adaptive unstructured road detection using close range stereo vision
EP4336466A2 (en) Method and apparatus for modeling object, storage medium, and vehicle control method
WO2023285254A1 (en) Generating a filtered point cloud and guiding a vehicle at least in part automatically
US20210350548A1 (en) Method for determining a relative movement using a digital image sequence

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
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20240228

Address after: 510000 No.8 Songgang street, Cencun, Tianhe District, Guangzhou City, Guangdong Province

Patentee after: GUANGZHOU XIAOPENG MOTORS TECHNOLOGY Co.,Ltd.

Country or region after: China

Address before: Room 46, room 406, No.1, Yichuang street, Zhongxin knowledge city, Huangpu District, Guangzhou City, Guangdong Province 510000

Patentee before: Guangzhou Xiaopeng Automatic Driving Technology Co.,Ltd.

Country or region before: China

TR01 Transfer of patent right