CN110780305B - Track cone detection and target point tracking method based on multi-line laser radar - Google Patents
Track cone detection and target point tracking method based on multi-line laser radar Download PDFInfo
- Publication number
- CN110780305B CN110780305B CN201910995353.2A CN201910995353A CN110780305B CN 110780305 B CN110780305 B CN 110780305B CN 201910995353 A CN201910995353 A CN 201910995353A CN 110780305 B CN110780305 B CN 110780305B
- Authority
- CN
- China
- Prior art keywords
- cone
- point cloud
- laser radar
- points
- point
- 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
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000001914 filtration Methods 0.000 claims abstract description 22
- 238000007619 statistical method Methods 0.000 claims abstract description 8
- 230000033001 locomotion Effects 0.000 claims abstract description 6
- 238000012545 processing Methods 0.000 claims description 10
- 238000009434 installation Methods 0.000 claims description 9
- 230000001133 acceleration Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000004804 winding Methods 0.000 claims description 5
- 230000005484 gravity Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 abstract description 4
- 230000001351 cycling effect Effects 0.000 abstract description 2
- 230000011218 segmentation Effects 0.000 abstract description 2
- 238000005286 illumination Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 4
- 238000012552 review Methods 0.000 description 4
- 230000004438 eyesight Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000004870 electrical engineering Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
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
- 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
-
- 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
- G01S17/66—Tracking systems using electromagnetic waves other than radio waves
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0257—Control of position or course in two dimensions specially adapted to land vehicles using a radar
-
- 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
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Electromagnetism (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Aviation & Aerospace Engineering (AREA)
- Automation & Control Theory (AREA)
- Optical Radar Systems And Details Thereof (AREA)
Abstract
The invention discloses a racetrack cone detection and target point tracking method based on a multi-line laser radar, which comprises the following steps: 1) Reading laser radar point cloud data; 2) Performing direct filtering on the laser radar point cloud data; 3) Eliminating interference of ground point cloud data on cone detection; 4) Screening out point cloud clusters of the cone barrel; 5) Carrying out statistical analysis on the point cloud clusters obtained by clustering, setting a maximum standard deviation threshold according to the characteristics of the actual size of the cone, and screening out the cone; 6) Acquiring the coordinates of the central point of the point cloud cluster; 7) Calculating average values of coordinates of central points of the cone barrels on the left side and the right side of the laser radar to obtain the closest target point in the current state of the central point of the cone barrel; 8) And (5) cycling the steps to obtain the latest target point. According to the method, the real-time filtering, segmentation, clustering and the like of the laser radar point cloud are carried out, so that the movement of the vehicle towards the target point is controlled continuously, and the track cone detection and the target point tracking based on the multi-line laser radar are finally realized.
Description
Technical Field
The invention relates to the field of environmental perception of equation-driven racing car, in particular to a track cone detection and target point tracking method based on multi-line laser radar.
Background
The college student's unmanned equation is a design and manufacturing race of unmanned racing vehicles attended by students in the university or the related profession of the automobile in the university. The event is known as a "cradling of an automobile engineer". In this event, multi-line lidar is commonly used by various unmanned racing teams as an important sensor for unmanned environmental awareness systems.
In the event, unmanned racing vehicles of the racing motorcade need to finish dynamic racing projects such as a linear acceleration project, an 8-shaped winding project, a high-speed tracking project and the like. The different racetracks are marked by cones of fixed dimensions (20 x 30 cm) according to the different racetrack shapes. According to the requirements of the event rules, before the racing car performs dynamic racing, survey and mapping of the racing track are not allowed, namely the unmanned racing car cannot acquire a map of the racing track to be completed in advance. Therefore, the cone barrel is an important mark for effectively identifying the track by the unmanned system, and the vehicle-mounted sensor is required to be fully utilized for detecting the boundary and the travelable area of the track in real time. The track cone detection and target point tracking method based on the multi-line laser radar is mainly applied to cone identification of the event, and can be widely applied to other similar scenes, such as parking lot cone detection in an autonomous parking environment.
The unmanned sensing sensor comprises a camera, a laser radar, a GPS inertial navigation sensor and other sensors, and the unmanned sensing sensor is mainly used for cone detection and target tracking method description aiming at the laser radar sensor. The lidar is classified into MEMS type lidar, flash type lidar, phased array lidar and mechanical rotation type lidar according to the scanning manner. Different types of laser radars have different scanning modes, and the manufacturing cost, the laser data processing mode and the application scene are different. The laser radars can be classified into single-line organization radars and multi-line laser radars according to the line number, and the single-line laser radars can only scan in a plane mode and are mainly applied to service robots such as sweeping robots; according to the multi-line laser radar, three-dimensional scanning is realized on space by different laser beams according to the density degree of the line number, the higher the laser line number is, the denser the scanned laser point cloud is, the more obvious the characteristic of the shape and the size of the object is, the larger the data volume is, and the more expensive the price is. Compared with a vision sensor such as a monocular camera, the laser radar has the advantages of being capable of acquiring high-precision depth information, target three-dimensional size information, being not easily influenced by illumination conditions and the like, and in the literature [ Ma Dianbo ], unmanned automobile environment sensing technology review [ J ]. Automobile and driving maintenance.2017 ], [ Wang Yifan ], automatic driving automobile sensing system key technology review [ J ]. Automobile electric appliance 2016 (12): 12-16 ] and [ Wang Yifan ], automatic driving automobile sensing system key technology review [ J ]. Automobile electric appliance 2016 (12): 12-16 ], the point cloud data scanned by the laser radar are sparse, and the defects that visual target texture and color information cannot be acquired are also overcome.
For target detection, currently, the mainstream target detection algorithm in the industry is mainly a visual recognition target detection algorithm designed for a visual sensor, and the target detection algorithm comprises a traditional image processing method and a target detection algorithm based on machine learning. With the rapid development of intelligent driving automobiles, the laser radar is gradually valued and applied by a plurality of automatic driving practitioners, and more people are put into the research of target detection algorithms based on laser radar sensors, wherein the method comprises a method for directly filtering and clustering laser point clouds so as to realize target detection and a point cloud target detection method based on deep learning. The method has the advantages that a plurality of point cloud data sets are required to be marked and trained, high requirements are provided for hardware computing power for processing data, and the implementation difficulty is high, so that the method mainly utilizes data processing methods such as point cloud filtering, clustering and the like to realize the target detection of the racetrack cone barrel.
For unmanned racing vehicles in the application scenario described above, [ Tang Zhiwei ] review of vision-based unmanned automobile research [ J ]]Automation of the manufacturing industry 2016 (08): 134-136.][Dhall, Ankit et al. “Real-time 3D Traffic Cone Detection for Autonomous Driving.” 2019 IEEE Intelligent Vehicles Symposium (IV) (2019): 494-501.]And [ Panagiotaki E An Efficient Track Detection and Mapping System for Autonomous Driving Race car [ D ]]. Department of Information Technology and Electrical Engineering, 2017.]In the literature, the acquisition of accurate target position information through a sensing system under the condition of too strong or too weak illumination intensity is also mentioned, and is a primary premise for ensuring decision planning of an unmanned system and stable operation of whole vehicle control. Therefore, in order to get rid of the influence of illumination conditions on an environment sensing system, the invention takes the multi-line laser radar as a research object and only depends on laser point cloud sensing dataAnd (3) researching a target detection and target point tracking algorithm on the racetrack cone. The cone barrel detection method is not easily affected by illumination conditions, does not need to carry out complicated data labeling, and can meet the requirement of real-time performance.
Disclosure of Invention
The technical problem to be solved by the invention is to realize unmanned racetrack cone detection and target tracking by only relying on laser radar point cloud data, and the problem that a visual perception system is susceptible to the influence of illumination intensity to fail is solved.
The invention is realized at least by one of the following technical schemes.
A racetrack cone detection and target point tracking method based on a multi-line laser radar comprises the following steps:
1) Reading laser radar point cloud data under an ROS robot operating system;
2) Adopting a direct filter to carry out direct filtering on the laser radar point cloud data according to different track scenes;
3) Adopting a random sampling consistency algorithm to eliminate interference of ground point cloud data on cone detection;
4) A Euclidean cluster extraction algorithm (Euclidean Cluster Extraction) is adopted to primarily screen out the point cloud clusters of the cone barrel;
5) Statistical analysis is carried out on the point cloud clusters obtained by clustering by adopting a maximum standard deviation threshold method, a X, Y, Z-direction maximum standard deviation threshold is set according to the characteristics of the actual size of the cone, and the point cloud clusters meeting the conditions are screened to be regarded as the cone;
6) Carrying out statistical analysis on the detected cone barrels, finding one cone barrel on the left side and the right side closest to the laser radar, and obtaining the coordinates of the center point of the point cloud cluster;
7) Calculating average values of coordinates of central points of cone barrels on the left side and the right side of the laser radar to obtain nearest target points in the current state of central points of nearest cone barrels on the two sides of the laser radar, and controlling the trolley to advance towards the target points;
8) Looping through steps 1) to 7), the latest target point is continuously acquired and tracked.
Further, step 1) specifically includes installing ROS (Robot Operating System) a robot operating system on a computing platform equipped with a ubuntu16.04 operating system, configuring a laser radar driving package, starting a laser radar running node to collect point cloud data in real time, and according to track characteristics, determining that the size of a cone to be detected is a x b, wherein in order to enable the laser radar to scan on the cone to the greatest extent, the laser radar installation position is located below the front nose of the racing car, and the installation height ground clearance is b/2.
Further, step 2) includes, before starting detection, presetting a track scene to be detected, including a 75-meter linear acceleration track, an 8-winding loop track and a high-speed tracking track; and setting detection ranges of the laser radars for different raceways, and filtering out point cloud data outside the ranges.
Further, the step 3) includes setting a plane filtering threshold to be a/5 by using a random sampling coincidence algorithm, judging the maximum deviation distance of points in the normal direction of the point cloud clusters extracted by the algorithm, regarding the point cloud clusters with the maximum deviation distance larger than a/5 as not being planes, regarding the point cloud clusters with the maximum deviation distance not larger than a/5 as the same plane in the normal direction of the point cloud clusters, and removing the point cloud clusters belonging to the planes in the current point cloud so as to achieve the purpose of filtering the ground point cloud data.
Further, step 4) specifically includes collecting point numbers on different cones of a racetrack by using a laser radar through analyzing point cloud data collected in advance, wherein the nearest cone point number is R, the most distant cone point number which can be swept is R, therefore, according to a statistical rule of collected data for a plurality of times, setting a clustering minimum point number as R points, a clustering maximum point number as R points, setting a maximum distance of searching two points in a clustering process as L, carrying out point cloud searching by using a KD tree, and dividing point clouds meeting a clustering condition, namely points within a range of (R, R) into a plurality of point cloud clusters; counting the number of points of each point cloud cluster, and respectively calculating the average value of X, Y, Z coordinates of each point cloud cluster to serve as the point cloud center of gravity point of the point cloud cluster, so as to replace the position of the point cloud cluster relative to the laser radar.
Further, step 5) specifically includes, according to the shape and size characteristics of the cone, setting the maximum standard deviation threshold value of the X, Y direction as Q, setting the cone height as b on the Z axis, setting the maximum standard deviation threshold value on the Z axis as Q, and considering the point cloud cluster as a qualified cone when the standard deviation of the X, Y, Z three-direction coordinate values of all the points in the point cloud cluster obtained by the clustering is counted to be smaller than the corresponding threshold value, that is, when the calculated X and Y-direction standard deviations of the point cloud cluster are smaller than Q and the Z-direction standard deviation is smaller than Q, because the cone is a rotating body around the central axis in the vertical direction.
Further, the step 6) specifically includes counting the number of points of the cone barrels detected in the step 5), and calculating an average Y coordinate of all points in each cone barrel; traversing all detected cones, finding the cone with the largest number of points in the cone with the average Y coordinate being more than 0, and considering the cone with the largest number of points in the cone with the average Y coordinate being less than 0 as the nearest cone on the left side of the laser radar, and similarly, finding the cone with the largest number of points in the cone with the average Y coordinate being less than 0, and considering the cone with the largest number of points in the cone as the nearest cone on the right side of the laser radar.
Further, the step 7) specifically includes calculating X, Y, Z coordinate averages of closest cone-barrel point cloud clusters from the left front and the right front of the laser radar, regarding the average value as center point coordinates of the closest cone-barrel of the laser radar, calculating midpoint coordinates of connecting lines of the left cone-barrel and the right cone-barrel of the laser radar, and taking the midpoint coordinates as target points in the current state, and the specific steps include calculating X, Y, Z coordinate averages of closest cone-barrel point cloud clusters from the left front and the right front of the laser radar obtained in the step 6) as spatial coordinate positions of the closest cone-barrel from the left front and the right front, and calculating center point coordinates of the connecting lines of the closest cone-barrel point cloud clusters as moving target points in the current state of the robot.
Further, the step 8) specifically includes performing the processing of the step 1) to the step 7) on the point cloud data collected by each frame of laser radar under the ROS robot operating system, and outputting a moving target point of the next frame of the robot after the processing of each frame of data is finished, so as to control the robot to move towards the target point, and at the same time, performing the processing of the step 1) to the step 7) on the data collected by the next frame of laser radar, so as to achieve the purposes of detecting the cone barrel in real time and tracking the target point.
Further, the laser radar is a 16-line laser radar.
Compared with the prior art, the invention has the beneficial effects that:
1. the cone barrel target detection method only depends on a laser radar sensor to perform operations such as filtering, segmentation, clustering and the like on the original data acquired by the laser radar so as to perform cone barrel detection and positioning. Compared with cone barrel detection based on stereoscopic vision, the method is not easily affected by illumination conditions, and has high ranging precision and high instantaneity.
2. The cone barrel target detection and target point tracking method can directly output the coordinate position of the target point relative to the vehicle, can adapt to different heights of laser radar installation by adjusting filtering parameters, is applied to various different types of vehicle chassis, and has strong adaptability.
Drawings
Fig. 1 is a schematic flow chart of a track cone detection and target point tracking method based on a multi-line laser radar in the embodiment;
FIG. 2 is a track diagram of the present embodiment;
wherein: 1-cone barrel and 2-trolley.
Detailed Description
The following describes the object of the present invention in further detail through specific embodiments, in order to reduce the cost and facilitate the experimental test, the present invention uses the experimental trolley to simulate the real application scene of the racing car, and uses the same installation position and installation height as the real car to simulate the data acquisition mode of the laser radar installed on the real racing car, and the embodiments cannot be described herein in detail, but the embodiments of the present invention are not limited to the following embodiments.
As shown in fig. 1, a track cone detection and target point tracking method based on multi-line laser radar comprises the following steps:
1. reading laser radar point cloud data under an ROS robot operating system; specifically, the step of reading laser radar point cloud data under the ROS robot operating system comprises the following steps: and installing an ROS robot operating system on a computing platform provided with a ubuntu16.04 operating system, configuring a laser radar driving package, and starting a laser radar operation node to acquire point cloud data in real time. The laser radar and algorithm operation platform, namely a notebook computer, is fixed on the experimental trolley in the embodiment, the size of the cone 1 to be detected is 20 x 30cm according to the track characteristics, the height of the cone 1 relative to the ground is lower, and in order to enable the laser radar point cloud to scan on the cone 1 to the greatest extent, the laser radar installation position is located on the forefront plane of the experimental trolley 2, and the installation height is 10cm from the ground clearance.
2. And adopting a direct filtering algorithm to carry out direct filtering on the laser radar point cloud data according to different track scenes, removing redundant point cloud data far away from the track, and reducing subsequent calculated amount.
Specifically, a track scene to be detected is preset before detection is started, and the track scene comprises a 75-meter linear acceleration track, an 8-winding loop track, a high-speed tracking track and the like. Setting a straight-pass filtering range for a 75-meter linear acceleration racing track: the range of X direction (0, 100) m right in front of the laser radar, the range of Y direction (-3, +3) m at the side of the laser radar, the range of (-0.5,0.3) m in the vertical Z direction of the laser radar, and filtering out the point cloud data outside the range. Aiming at the 8-shaped winding circular racetrack, a through filtering range is set, an isosceles trapezoid which takes a Y axis as a lower bottom edge, is 6m long and 10 m high is arranged on a laser radar XY plane, the included angle of left and right oblique sides is 90 degrees, and point cloud data outside the range is filtered in the vertical Z direction (-0.5,0.3) m range of the laser radar. Aiming at a high-speed tracking track, a straight-through filtering range is set, an isosceles trapezoid with a Y axis as a lower bottom edge, a length of 6m and a height of 20m and a left and right bevel edge included angle of 90 degrees is set on a laser radar XY plane, an X-direction (0, 20) meter range is right in front of the laser radar, a Z-direction (-0.5,0.3) meter range of the laser radar is vertical, and point cloud data outside the range is filtered. Fig. 2 is a straight acceleration track.
3. And a random sampling consensus algorithm (RANSAC algorithm) is adopted to eliminate interference of ground point cloud data on cone detection.
The method mainly aims at filtering point cloud data of a laser radar scanned on the ground and eliminating interference of the ground point cloud on cone detection. Filtering the ground point cloud is one link in the RANSAC algorithm, which can choose to only reserve Ping Miandian cloud, or can choose to only filter Ping Miandian cloud.
4. And (3) primarily screening out point cloud clusters which are possibly cone barrels by adopting a Euclidean clustering algorithm. Specifically, by analyzing a point cloud data packet acquired in advance, the laser radar used in the invention is a 16-line laser radar, and 16 wire bundles scanned during working are uniformly distributed in an included angle of 30 degrees by taking a horizontal plane as a symmetry plane, and the distance which can be scanned at the maximum is 150m. The number of points scanned on different cones of the track changes with distance, the number of the nearest cones is close to 120, and the number of the most distant cones is only 2. Therefore, according to the statistical rule of the data collected for many times, the minimum point number of the clustering is set to be 2 points, the maximum point number of the clustering is set to be 120 points, the maximum distance between two points searched in the clustering process is set to be 0.3m, the KD tree is adopted for carrying out point cloud searching, and the point cloud meeting the clustering condition is divided into a plurality of point cloud clusters. And counting the number of points of each point cloud cluster, and respectively calculating the average value of X, Y, Z coordinates of each point cloud cluster to serve as the point cloud gravity center point of the point cloud cluster, wherein the gravity center point is not the geometric center point of the point cloud cluster, but can be used for approximately replacing the position of the point cloud cluster relative to a laser radar.
5. And carrying out statistical analysis on the point cloud clusters obtained by clustering by adopting a maximum standard deviation threshold algorithm, setting a maximum standard deviation threshold in the X, Y, Z direction according to the characteristics of the actual size of the cone, and screening out the point cloud clusters meeting the conditions to be regarded as the cone.
Specifically, the statistical analysis is performed on the point cloud clusters obtained by clustering by adopting a maximum standard deviation threshold algorithm, a maximum standard deviation threshold in the X, Y, Z direction is set according to the characteristics of the actual size of the cone, and the point cloud clusters meeting the conditions are screened to be regarded as the cone, comprising the following steps: according to the shape and size characteristics of the cone, since the cone can be regarded as a rotating body around the central axis in the vertical direction, the characteristics are consistent in the direction X, Y, the maximum standard deviation threshold is set to be 0.08m, the cone height is 30cm in the Z axis, the maximum standard deviation threshold is set to be 0.15m, and when the standard deviation of X, Y, Z three direction coordinate values of all points in the point cloud cluster obtained by clustering is counted to be smaller than the corresponding threshold, the point cloud cluster is regarded as a cone meeting the conditions.
6. And carrying out statistical analysis on the detected cone barrels, finding one cone barrel on the left side and the right side closest to the laser radar, and obtaining the coordinates of the center point of the point cloud cluster. Specifically, counting the number of points of the cone detected in the previous step, and calculating the average Y coordinate mean_Y of all points in each cone. Traversing all detected cone barrels, finding the cone barrel with the average Y coordinate larger than 0, namely the cone barrel with mean_Y >0 and the largest point number, regarding the cone barrel as the nearest cone barrel on the left side of the laser radar, and discarding the rest point cloud clusters. Similarly, finding the cone with the average Y coordinate less than 0, i.e., mean_y <0, with the largest number of points in the cone is considered the closest cone to the right of the lidar.
7. And calculating an average value of coordinates of central points of the left cone barrel and the right cone barrel to obtain a nearest target in the current state of the central point of the left nearest cone barrel and the right nearest cone barrel, and controlling the trolley to advance towards the target. Specifically, the average value of X, Y, Z coordinates is calculated for the point cloud clusters of the nearest cones at the left front and the right front of the laser radar obtained in the last step respectively, and is used as the space coordinate position of the nearest cones at the left front and the right front, and the central point coordinate of the connecting line is calculated and is used as the motion target point of the robot in the current state.
8. And circulating the steps to continuously acquire and track the latest target point. And calculating a next movement target point in the current state of the vehicle according to the steps, adjusting the vehicle orientation angle according to the coordinate position of the next movement target point relative to the vehicle, and controlling the vehicle to move towards the target point. Meanwhile, the laser radar acquires the next frame of point cloud data, starts a new point cloud data processing step, obtains a next movement target point, and controls the vehicle to continue to advance. And continuously cycling the steps of point cloud processing and vehicle control above the ROS robot operating system, and realizing track cone detection and target point tracking based on the multi-line laser radar.
The above description is only of the preferred embodiments of the present invention, but the protection scope of the present invention is not limited thereto, and any person skilled in the art can make equivalent substitutions or modifications according to the technical solution of the present invention and the inventive concept thereof within the scope of the present invention disclosed in the present invention.
Claims (10)
1. The track cone barrel detection and target point tracking method based on the multi-line laser radar is characterized by comprising the following steps of:
1) Reading laser radar point cloud data under an ROS robot operating system;
2) Adopting a direct filter to carry out direct filtering on the laser radar point cloud data according to different track scenes;
3) Adopting a random sampling consistency algorithm to eliminate interference of ground point cloud data on cone detection;
4) A Euclidean clustering extraction algorithm is adopted to primarily screen out the point cloud clusters of the cone barrel;
5) Statistical analysis is carried out on the point cloud clusters obtained by clustering by adopting a maximum standard deviation threshold method, a X, Y, Z-direction maximum standard deviation threshold is set according to the characteristics of the actual size of the cone, and the point cloud clusters meeting the conditions are screened to be regarded as the cone;
6) Carrying out statistical analysis on the detected cone barrels, finding one cone barrel on the left side and the right side closest to the laser radar, and obtaining the coordinates of the center point of the point cloud cluster;
7) Calculating average values of coordinates of central points of cone barrels on the left side and the right side of the laser radar to obtain nearest target points in the current state of central points of nearest cone barrels on the two sides of the laser radar, and controlling the trolley to advance towards the target points;
8) Looping through steps 1) to 7), the latest target point is continuously acquired and tracked.
2. The method for detecting and tracking target points on a racetrack cone based on multi-line lidar according to claim 1, wherein step 1) specifically comprises the steps of installing a ROS (Robot Operating System) robot operating system on a computing platform equipped with a ubuntu16.04 operating system, configuring a lidar driving package, starting a lidar operating node to collect point cloud data in real time, and according to racetrack characteristics, the size of the detected cone is a x b, and in order to enable the lidar to scan on the cone to the greatest extent, the installation position of the lidar is located below the front nose of the racetrack, and the installation height ground clearance is b/2.
3. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: step 2) presetting a track scene to be detected before starting detection, wherein the track scene comprises a 75-meter linear acceleration track, an 8-shaped winding track and a high-speed tracking track; and setting detection ranges of the laser radars for different raceways, and filtering out point cloud data outside the ranges.
4. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: and 3) setting a plane filtering threshold value as a/5 by utilizing a random sampling coincidence algorithm, judging the maximum deviation distance of points in the normal direction of the point cloud clusters extracted by the algorithm, regarding the point cloud clusters with the maximum deviation distance larger than a/5 as not being planes, regarding the point cloud clusters with the maximum deviation distance not larger than a/5 as the same plane in the normal direction of the point cloud clusters, and removing the point cloud clusters belonging to the planes in the current point cloud so as to achieve the purpose of filtering the ground point cloud data.
5. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: step 4) specifically, acquiring the number of points on different cones of a racetrack by adopting a laser radar through analyzing point cloud data acquired in advance, wherein the number of points of the nearest cone is R, the number of points of the cone which can be scanned furthest is R, therefore, according to the statistical rule of acquired data for a plurality of times, the minimum number of clustering is R points, the maximum number of clustering is R points, the maximum distance of searching two points in the clustering process is L, the point cloud searching is carried out by adopting a KD tree, and the point cloud which meets the clustering condition, namely the points within the range of (R and R), is divided into a plurality of point cloud clusters; counting the number of points of each point cloud cluster, and respectively calculating the average value of X, Y, Z coordinates of each point cloud cluster to serve as the point cloud center of gravity point of the point cloud cluster, so as to replace the position of the point cloud cluster relative to the laser radar.
6. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: step 5) specifically includes setting the maximum standard deviation threshold value in the X, Y direction as Q, setting the cone height as b in the Z axis, setting the maximum standard deviation threshold value in the Z axis as Q, and considering the point cloud cluster as a cone meeting the conditions when the standard deviation of the X, Y, Z three direction coordinate values of all the points in the point cloud cluster obtained by the clustering is counted to be smaller than the corresponding threshold value, namely, when the standard deviation of the X and Y directions of the point cloud cluster is calculated to be smaller than Q and the standard deviation of the Z direction is smaller than Q, because the cone is a rotating body around the central axis in the vertical direction according to the shape and size characteristics of the cone.
7. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: the step 6) specifically includes counting the number of points of the cone barrels detected in the step 5), and calculating the average Y coordinates of all points in each cone barrel; traversing all detected cones, finding the cone with the largest number of points in the cone with the average Y coordinate being more than 0, and considering the cone with the largest number of points in the cone with the average Y coordinate being less than 0 as the nearest cone on the left side of the laser radar, and similarly, finding the cone with the largest number of points in the cone with the average Y coordinate being less than 0, and considering the cone with the largest number of points in the cone as the nearest cone on the right side of the laser radar.
8. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: the step 7) specifically includes calculating X, Y, Z coordinate average values of closest cone barrel point cloud clusters in the left front and the right front of the laser radar respectively, regarding the average values as center point coordinates of left and right closest cone barrels of the laser radar, calculating midpoint coordinates of connecting lines of the left and right cone barrels of the laser radar, and taking the midpoint coordinates as target points in the current state, and the specific steps include calculating X, Y, Z coordinate average values of closest cone barrel point cloud clusters in the left front and the right front of the laser radar obtained in the step 6) respectively, taking the average values as spatial coordinate positions of the left front closest cone barrel and the right front closest cone barrel, and calculating center point coordinates of the connecting lines of the closest cone barrel point average values as motion target points in the current state of the robot.
9. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: step 8) specifically includes that under the operation system of the ROS robot, the processing of step 1) to step 7) is carried out on the point cloud data collected by each frame of laser radar, a moving target point of the next frame of the robot is output after the processing of each frame of data is finished, the robot is controlled to move towards the target point, and meanwhile, the processing of step 1) to step 7) is carried out on the data collected by the next frame of laser radar, so that the purposes of detecting a cone barrel in real time and tracking the target point are achieved.
10. The method for detecting and tracking the target point of the racetrack cone based on the multi-line laser radar according to claim 1, wherein the method comprises the following steps: the laser radar is a 16-line laser radar.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910995353.2A CN110780305B (en) | 2019-10-18 | 2019-10-18 | Track cone detection and target point tracking method based on multi-line laser radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910995353.2A CN110780305B (en) | 2019-10-18 | 2019-10-18 | Track cone detection and target point tracking method based on multi-line laser radar |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110780305A CN110780305A (en) | 2020-02-11 |
| CN110780305B true CN110780305B (en) | 2023-04-21 |
Family
ID=69385978
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910995353.2A Active CN110780305B (en) | 2019-10-18 | 2019-10-18 | Track cone detection and target point tracking method based on multi-line laser radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110780305B (en) |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10788835B2 (en) | 2017-11-13 | 2020-09-29 | Raven Industries, Inc. | Safety system for autonomous operation of off-road and agricultural vehicles using machine learning for detection and identification of obstacles |
| CN111308499A (en) * | 2020-03-09 | 2020-06-19 | 中振同辂(江苏)机器人有限公司 | Obstacle detection method based on multi-line laser radar |
| CN111487641B (en) * | 2020-03-19 | 2022-04-22 | 福瑞泰克智能系统有限公司 | Method and device for detecting object by using laser radar, electronic equipment and storage medium |
| EP4143605A4 (en) * | 2020-04-28 | 2023-10-25 | Raven Industries, Inc. | Object detection and tracking for automated operation of vehicles and machinery |
| CN111723716B (en) * | 2020-06-11 | 2024-03-08 | 深圳地平线机器人科技有限公司 | Method, device, system, medium and electronic equipment for determining target object orientation |
| CN111709355B (en) * | 2020-06-12 | 2023-08-29 | 阿波罗智联(北京)科技有限公司 | Method and device for identifying target area, electronic equipment and road side equipment |
| CN112101128B (en) * | 2020-08-21 | 2021-06-22 | 东南大学 | Unmanned formula racing car perception planning method based on multi-sensor information fusion |
| CN112294197A (en) * | 2020-11-04 | 2021-02-02 | 深圳市普森斯科技有限公司 | Sweeping control method of sweeper, electronic device and storage medium |
| CN112415538A (en) * | 2020-11-16 | 2021-02-26 | 重庆长安汽车股份有限公司 | Planning method and system for automatic driving vehicle to detour conical cylinder and vehicle |
| CN112731925B (en) * | 2020-12-21 | 2024-03-15 | 浙江科技学院 | Cone barrel identification and path planning and control method for formula car |
| CN112578673B (en) * | 2020-12-25 | 2022-08-02 | 浙江科技学院 | Perception decision and tracking control method for multi-sensor fusion of formula-free racing car |
| CN112883820B (en) * | 2021-01-26 | 2024-04-19 | 上海应用技术大学 | Road target 3D detection method and system based on laser radar point cloud |
| CN112884687B (en) * | 2021-01-29 | 2022-02-11 | 郑州信大云筑工程科技有限公司 | Mapping laser radar scanning strategy control method and system based on artificial intelligence |
| CN113778078A (en) * | 2021-03-05 | 2021-12-10 | 北京京东乾石科技有限公司 | Positioning information generation method and device, electronic equipment and computer readable medium |
| CN115049696A (en) * | 2021-03-08 | 2022-09-13 | 北京金茂绿建科技有限公司 | Personnel monitoring method and device based on radar data |
| CN113095324A (en) * | 2021-04-28 | 2021-07-09 | 合肥工业大学 | Classification and distance measurement method and system for cone barrel |
| CN113325428A (en) * | 2021-04-29 | 2021-08-31 | 中国科学院沈阳自动化研究所 | Distance and course measuring method based on laser radar |
| CN113343840B (en) * | 2021-06-02 | 2022-03-08 | 合肥泰瑞数创科技有限公司 | Object identification method and device based on three-dimensional point cloud |
| CN113655498B (en) * | 2021-08-10 | 2023-07-18 | 合肥工业大学 | Method and system for extracting cone barrel information in racetrack based on laser radar |
| CN114387305A (en) * | 2021-12-30 | 2022-04-22 | 驭势科技（北京）有限公司 | Trailer queue tracking method and device, computer equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018205119A1 (en) * | 2017-05-09 | 2018-11-15 | 深圳市速腾聚创科技有限公司 | Roadside detection method and system based on laser radar scanning |
| CN108932736A (en) * | 2018-05-30 | 2018-12-04 | 南昌大学 | Two-dimensional laser radar Processing Method of Point-clouds and dynamic robot pose calibration method |
| CN109188459A (en) * | 2018-08-29 | 2019-01-11 | 东南大学 | A kind of small obstacle recognition method in ramp based on multi-line laser radar |
| CN110018489A (en) * | 2019-04-25 | 2019-07-16 | 上海蔚来汽车有限公司 | Target tracking method, device and controller and storage medium based on laser radar |
| CN110275153A (en) * | 2019-07-05 | 2019-09-24 | 上海大学 | A kind of waterborne target detection and tracking based on laser radar |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107945198B (en) * | 2016-10-13 | 2021-02-23 | 北京百度网讯科技有限公司 | Method and device for marking point cloud data |
-
2019
- 2019-10-18 CN CN201910995353.2A patent/CN110780305B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018205119A1 (en) * | 2017-05-09 | 2018-11-15 | 深圳市速腾聚创科技有限公司 | Roadside detection method and system based on laser radar scanning |
| CN108932736A (en) * | 2018-05-30 | 2018-12-04 | 南昌大学 | Two-dimensional laser radar Processing Method of Point-clouds and dynamic robot pose calibration method |
| CN109188459A (en) * | 2018-08-29 | 2019-01-11 | 东南大学 | A kind of small obstacle recognition method in ramp based on multi-line laser radar |
| CN110018489A (en) * | 2019-04-25 | 2019-07-16 | 上海蔚来汽车有限公司 | Target tracking method, device and controller and storage medium based on laser radar |
| CN110275153A (en) * | 2019-07-05 | 2019-09-24 | 上海大学 | A kind of waterborne target detection and tracking based on laser radar |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110780305A (en) | 2020-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110780305B (en) | Track cone detection and target point tracking method based on multi-line laser radar | |
| CN111958592B (en) | Image semantic analysis system and method for transformer substation inspection robot | |
| CN110320504B (en) | Unstructured road detection method based on laser radar point cloud statistical geometric model | |
| CN110531376B (en) | Obstacle detection and tracking method for port unmanned vehicle | |
| CA2950791C (en) | Binocular visual navigation system and method based on power robot | |
| CN108549084B (en) | Target detection and attitude estimation method based on sparse two-dimensional laser radar | |
| CN109816811A (en) | A kind of nature driving data acquisition device | |
| CN106054900A (en) | Temporary robot obstacle avoidance method based on depth camera | |
| KR20180044279A (en) | System and method for depth map sampling | |
| CN105404844A (en) | Road boundary detection method based on multi-line laser radar | |
| CN104657968B (en) | Automatic vehicle-mounted three-dimensional laser point cloud facade classification and outline extraction method | |
| CN107356933A (en) | A kind of unstructured road detection method based on four line laser radars | |
| CN114051628B (en) | Method and device for determining target object point cloud set | |
| CN115861968A (en) | Dynamic obstacle removing method based on real-time point cloud data | |
| CN111766608A (en) | Environmental perception system based on laser radar | |
| CN113820682A (en) | Target detection method and device based on millimeter wave radar | |
| CN116508071A (en) | System and method for annotating automotive radar data | |
| Yang et al. | Car-Following Behavior Based on LiDAR Trajectory Data at Urban Intersections | |
| CN113312403B (en) | Map acquisition method and device, electronic equipment and storage medium | |
| CN116913081A (en) | Vehicle queuing length detection method based on roadside laser radar | |
| CN116206286A (en) | Obstacle detection method, device, equipment and medium under high-speed road condition | |
| Yang et al. | Extraction of straight field roads between farmlands based on agricultural vehicle-mounted LiDAR | |
| CN116087988A (en) | Multi-line laser radar obstacle detection method | |
| CN115421160A (en) | Road edge detection method, device, equipment, vehicle and storage medium | |
| CN112883846A (en) | Three-dimensional data acquisition imaging system for detecting vehicle front target |
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 |