CN110780305B - Track cone detection and target point tracking method based on multi-line laser radar - Google Patents

Abstract

The invention discloses a multi-line lidar-based racetrack cone detection and target point tracking method, comprising the following steps: 1) reading lidar point cloud data; 2) performing straight-through filtering on the lidar point cloud data; 3 ) Eliminate the interference of the ground point cloud data on the cone bucket detection; 4) Screen out the point cloud clusters of the cone bucket; 5) Perform statistical analysis on the point cloud clusters obtained by clustering, and set the maximum standard deviation according to the characteristics of the actual size of the cone bucket Threshold, filter out the cone bucket; 6) Obtain the coordinates of the center point of the point cloud cluster; 7) Calculate the average value of the center point coordinates of the cone buckets on the left and right sides of the lidar, and get the center point of the cone bucket as the nearest target in the current state 8) Repeat the above steps to obtain the latest target point. The invention continuously controls the movement of the vehicle towards the target point through real-time filtering, segmentation, clustering and other processing of the laser radar point cloud, and finally realizes the cone detection and target point tracking of the track based on the multi-line laser radar.

Description

A racetrack cone detection and target point tracking method based on multi-line lidar

technical field

The invention relates to the field of environmental perception of driverless formula racing cars, in particular to a method for detecting cones on a track and tracking a target point based on a multi-line laser radar.

Background technique

The Chinese College Student Driverless Formula Contest (English test: FSAC) is a design and manufacturing competition for driverless racing cars organized by students majoring in automotive engineering or automotive related majors in colleges and universities. This event is known as "the cradle of automotive engineers". In this event, various unmanned racing teams generally used multi-line lidar as an important sensor for the unmanned driving environment perception system.

In this event, the unmanned racing cars of the participating teams need to complete dynamic competitions such as linear acceleration, figure-of-eight circling, and high-speed tracking. Different tracks are marked by cones of fixed size (20*20*30cm) according to different track shapes. According to the requirements of the competition rules, before the dynamic racing of the racing car, it is not allowed to survey and map the track, that is, the unmanned racing car cannot obtain the track map to be completed in advance. Therefore, the cone is an important sign for the unmanned driving system to effectively identify the track, and it is necessary to make full use of the on-board sensors to detect the track boundary and drivable area in real time. The track cone detection and target point tracking method based on multi-line laser radar in the present invention is mainly applied to the cone recognition of this event, and can be extended to other similar scenarios, such as parking cones in the autonomous parking environment. barrel detection etc.

Unmanned driving perception sensors include various sensors such as cameras, laser radars, and GPS inertial navigation. Laser radar can be classified into MEMS laser radar, Flash laser radar, phased array laser radar and mechanical rotary laser radar according to the scanning method. Different types of lidar have different scanning methods, and the manufacturing costs, laser data processing methods, and application scenarios are all different. According to the number of lines, lidar can be divided into single-line mechanism radar and multi-line lidar. Single-line lidar can only scan in a plane, and is mainly used in service robots such as sweeping robots; multi-line lidar is based on the density of lines. To a certain extent, different laser line beams realize three-dimensional scanning in space according to a certain angle. The higher the number of laser lines, the denser the scanned laser point cloud, the more obvious the target shape and size characteristics, the greater the amount of data, and the lower the price. more expensive. Compared with visual sensors such as monocular cameras, lidar has the advantages of being able to obtain high-precision depth information, target three-dimensional size information, and not easily affected by lighting conditions. and Driving Maintenance. 2017.], [Wang Yifan. A Review of Key Technologies for Autonomous Vehicle Perception Systems[J]. Auto Electrical Appliances. 2016(12):12-16.] and [Wang Yifan. A Review of Key Technologies for Autonomous Vehicle Perception Systems[J]. ]. Automobile Electrical Appliances. 2016(12):12-16.], because the point cloud data scanned by lidar is relatively sparse, it also has the disadvantage of being unable to obtain visual target texture and color information.

For target detection, the current mainstream target detection algorithms in the industry are mainly visual recognition target detection algorithms designed for visual sensors, including traditional image processing methods and machine learning-based target detection algorithms. With the rapid development of intelligent driving vehicles, lidar has gradually attracted the attention and application of many autonomous driving practitioners, and more and more people have invested in the research of target detection algorithms based on lidar sensors, including direct filtering of laser point clouds , clustering and other data processing methods to realize the target detection method and the point cloud target detection method based on deep learning. The latter needs to mark and train many point cloud data sets, which has high requirements for hardware computing power to process data and is difficult to implement. Therefore, the present invention mainly uses data processing methods such as point cloud filtering and clustering to realize the track Cone bucket object detection.

For the unmanned racing car in the above application scenarios, [Tang Zhiwei. Vision-based unmanned vehicle research review [J]. Manufacturing Automation. 2016(08): 134-136.][Dhall, Ankit et al al. “Real-time 3D Traffic Cone Detection for Autonomous Driving.” 2019 IEEE Intelligent Vehicles Symposium (IV) (2019): 494-501.] and [Panagiotaki E. An Efficient TrackDetection and Mapping System for Autonomous Driving Race car[D ]. Department of Information Technology and Electrical Engineering, 2017.] It is also mentioned in the literature that when the light intensity is too strong or too weak, obtaining accurate target position information through the perception system is the key to ensure the decision-making planning and integration of the unmanned driving system. The first prerequisite for the stable operation of the vehicle control. Therefore, in order to get rid of the influence of lighting conditions on the environmental perception system, the present invention takes multi-line laser radar as the research object, and only relies on the laser point cloud perception data to conduct target detection and target point tracking algorithm research on the track cone barrel. The cone bucket detection method of the present invention is not easily affected by illumination conditions, does not need cumbersome data labeling, and can meet real-time requirements.

Contents of the invention

The technical problem to be solved by the present invention is to realize the cone detection and target tracking of the unmanned driving track only relying on the laser radar point cloud data, and get rid of the problem that the visual perception system is easily affected by the light intensity and becomes invalid.

The present invention is realized through at least one of the following technical solutions.

A racetrack cone detection and target point tracking method based on multi-line lidar, comprising the following steps:

1) Read the lidar point cloud data under the ROS robot operating system;

2) Use a pass-through filter to perform pass-through filtering on the lidar point cloud data according to different track scenarios;

3) Adopt the random sampling consensus algorithm to eliminate the interference of the ground point cloud data on the cone bucket detection;

4) Use the Euclidean Cluster Extraction algorithm (Euclidean Cluster Extraction) to initially screen out the point cloud clusters of the cone buckets;

5) Use the method of the maximum standard deviation threshold to perform statistical analysis on the point cloud clusters obtained by clustering. According to the characteristics of the actual size of the cone bucket, set the maximum standard deviation threshold in the X, Y, and Z directions to filter out qualified point clouds Clusters are considered cone buckets;

6) Statistically analyze the detected cone buckets, find one cone bucket on the left and right sides closest to the lidar, and obtain the coordinates of the center point of the point cloud cluster;

7) Calculate the average value of the center point coordinates of the cone barrels on the left and right sides of the laser radar, and obtain the center point of the nearest cone barrel on both sides of the laser radar as the nearest target point in the current state, and control the car to move towards the target point;

8) Cycle step 1) to step 7) to continuously acquire and track the latest target point.

Further, step 1) specifically includes installing the ROS (Robot Operating System) robot operating system on the computing platform equipped with the ubuntu16.04 operating system, configuring the lidar driver package, and starting the lidar running node for real-time point cloud data Acquisition, according to the characteristics of the track, the size of the cone to be detected is a*a*b. In order to make the lidar scan on the cone to the greatest extent, the installation position of the lidar is located under the front nose of the racing car. The altitude ground clearance is b/2.

Further, step 2) includes, before starting the detection, pre-setting the track scene to be detected, including a 75-meter linear acceleration track, a figure-eight circle track, and a high-speed tracking track; setting lasers for different tracks The detection range of the radar, and filter out the point cloud data outside the range.

Further, step 3) includes, using the random sampling consensus algorithm to set the plane filtering threshold to a/5, judging the maximum deviation distance of points in the normal direction of the point cloud cluster extracted by the algorithm, and setting the maximum deviation distance greater than a/5 The point cloud clusters are not considered to be planes, and the point cloud clusters within the range of the maximum deviation distance not greater than a/5 will be considered as the same plane in the normal direction of the point cloud clusters, and the point cloud clusters belonging to the plane will be removed from the current point cloud , in order to achieve the purpose of filtering ground point cloud data.

Further, step 4) is specifically to analyze the pre-collected point cloud data and use lidar to collect the number of points on different cones on the track. The nearest cone point is R, and the farthest cone point can be scanned Therefore, according to the statistical law of several times of data collection, set the minimum number of clustering points to be r points, the maximum number of clustering points to be R points, set the maximum distance between two points in the clustering process to be L, and use KD The tree performs point cloud search, and divides the point cloud that meets the clustering condition, that is, the number of points in the range of (r, R) into several point cloud clusters; counts the number of points in each point cloud cluster, and separates the points for each point cloud cluster Calculate the average value of the X, Y, and Z coordinates as the point cloud center of gravity of the point cloud cluster, which is used to replace the position of the point cloud cluster relative to the lidar.

Further, step 5) specifically includes, according to the shape and size characteristics of the cone barrel, since the cone barrel is a rotating body around the central axis in the vertical direction, the width in the X and Y directions is a, so the maximum in the X and Y directions The standard deviation threshold is set to Q, the height of the cone on the Z axis is b, and the maximum standard deviation threshold on the Z axis is set to q, when the X, Y, Z of all points in the point cloud cluster obtained by clustering are counted When the standard deviation of the coordinate values in the three directions is less than the corresponding threshold, that is, when the calculated standard deviation of the X and Y directions of the point cloud cluster is less than Q and the standard deviation of the Z direction is less than q, the point cloud cluster is considered to be a qualified cone. bucket.

Further, step 6) specifically includes, counting the points of the cone buckets detected in step 5), and calculating the average Y coordinate of all points in each cone bucket; traversing all the detected cone buckets, and finding that the average Y coordinate is greater than 0 The cone with the most points in the cone is considered to be the nearest cone on the left side of the lidar. Similarly, the cone with the most points found among the cones whose average Y coordinate is less than 0 is considered to be the nearest cone to the right of the lidar.

Further, step 7) specifically includes calculating the average values of the X, Y, and Z coordinates of the cone bucket point cloud clusters closest to the left front and right front of the lidar, and considering them as the center point coordinates of the nearest cone buckets on the left and right sides of the lidar, and Calculate the coordinates of the midpoint of the connecting line between the left and right cones of the lidar, and use it as the target point in the current state. The specific steps include, for the point cloud clusters of the nearest cones in the left front and right front of the lidar obtained in step 6), respectively Calculate the average value of the X, Y, and Z coordinates as the spatial coordinate positions of the nearest cone barrels in the left front and right front, and calculate the center point coordinates of their connections as the moving target point in the current state of the robot.

Further, step 8) specifically includes processing the point cloud data collected by each frame of lidar from step 1) to step 7) under the ROS robot operating system, and outputting the next step of the robot after each frame of data processing is completed. At the same time, the data collected by the lidar in the next frame is also processed from step 1) to step 7), so as to achieve real-time detection of the cone barrel and the realization of target point tracking. Purpose.

Further, the lidar is a 16-line lidar.

Compared with prior art, the beneficial effect of the present invention is:

1. The cone-barrel target detection method of the present invention only relies on the laser radar sensor to perform filtering, segmentation, clustering and other operations on the original data collected by the laser radar to detect and locate the cone-barrel. Compared with the cone detection based on stereo vision, it is not easily affected by lighting conditions, has higher ranging accuracy and stronger real-time performance.

2. The cone bucket target detection and target point tracking method of the present invention can directly output the coordinate position of the target point relative to the vehicle itself, and can adapt to different heights of laser radar installation by adjusting the filter parameters, and the algorithm migration can be applied to multiple Different types of vehicle chassis, strong adaptability.

Description of drawings

Fig. 1 is a schematic flow chart of a multi-line lidar-based racetrack cone detection and target point tracking method in this embodiment;

The track diagram of Fig. 2 present embodiment;

Among them: 1-cone barrel, 2-trolley.

Detailed ways

The purpose of the present invention will be further described in detail through specific examples below. In order to reduce costs and facilitate experimental testing, the present invention uses the experimental car 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 laser radar The data collection method installed on the real racing car, the embodiment can not be repeated here one by one, but the embodiment of the present invention is not limited to the following embodiment.

As shown in Figure 1, a racetrack cone detection and target point tracking method based on multi-line lidar includes the following steps:

1. Read the lidar point cloud data under the ROS robot operating system; specifically, the described reading of the lidar point cloud data under the ROS robot operating system includes the steps: On the platform, install the ROS robot operating system, configure the lidar driver package, and start the lidar running node for real-time collection of point cloud data. Fix the laser radar and the algorithm running platform, that is, the notebook computer, on the experimental car described in the embodiment. According to the characteristics of the track, the size of the cone barrel 1 to be detected is 20*20*30cm, and the height of the cone barrel 1 relative to the ground is low. In order to scan the laser radar point cloud on the cone barrel 1 to the greatest extent, the laser radar installation position in the embodiment of the present invention is located at the front plane of the experimental car 2, and the installation height and the ground clearance are 10 cm.

2. Adopt the straight-through filtering algorithm to perform straight-through filtering on the lidar point cloud data according to different track scenes, remove redundant point cloud data far away from the track, and reduce the amount of follow-up calculations.

Specifically, before starting the detection, the track scene to be detected is set in advance, including a 75-meter linear acceleration track, a figure-eight circle track, a high-speed tracking track, and the like. For the 75-meter linear acceleration track, set the through-filtering range: (0, 100) meters in the X direction directly in front of the lidar, (-3, +3) meters in the Y direction on the side of the lidar, and (-3, +3) meters in the vertical Z direction of the lidar ( -0.5, 0.3) meter range, filter out point cloud data outside this range. For the 8-shaped circle track, set the through-filtering range, and set the Y-axis as the bottom edge on the XY plane of the laser radar. The range of (-0.5, 0.3) meters in the vertical Z direction, filter out the point cloud data outside this range. For the high-speed tracking track, set the through-filtering range. Set the Y-axis as the bottom edge on the XY plane of the laser radar. It is 6m long, 20m high, and the angle between the left and right hypotenuses is 90 degrees. It is directly in front of the laser radar. The range of (0, 20) meters in the X direction, and the range of (-0.5, 0.3) meters in the vertical Z direction of the lidar, filter out the point cloud data outside this range. Figure 2 is a linear acceleration track.

3. Adopt the random sampling consensus algorithm (RANSAC algorithm) to eliminate the interference of the ground point cloud data on the cone bucket detection.

Set a plane filter for the point cloud after the through filter. Since the track ground is generally asphalt road surface, there are often some bumps and undulations on the road surface. The point cloud within a deviation distance of not more than 0.04m in one direction is considered to be the same plane. The main purpose of this step is to filter out the point cloud data scanned by the lidar on the ground and eliminate the interference of the ground point cloud on the cone bucket detection. Filtering the ground point cloud is a link in the RANSAC algorithm. The algorithm can choose to keep only the plane point cloud or filter out the plane point cloud only.

4. Use the Euclidean clustering algorithm to initially screen out point cloud clusters that may be cone buckets. Specifically, by analyzing the pre-collected point cloud data packets, the laser radar used in the present invention is a 16-line laser radar, and the 16 line beams that are scanned during operation are evenly distributed within a 30-degree angle with the horizontal plane as a symmetrical plane. The distance that can be scanned is 150m. The number of points swept on different cones on the track varies with the distance. The nearest cone has nearly 120 points, while the farthest cone has only 2 points. Therefore, according to the statistical law of multiple data collection, the minimum number of clustering points is set to 2 points, the maximum number of clustering points is 120 points, and the maximum distance between two points in the clustering process is set to 0.3m. Cloud search, which divides the point cloud meeting the clustering conditions into multiple point cloud clusters. Count the number of points in each point cloud cluster, and calculate the average value of X, Y, and Z coordinates for each point cloud cluster, as the point cloud center of gravity of the point cloud cluster, which is not the geometric center of the point cloud cluster points, but can be used to approximate the position of the point cloud cluster relative to the lidar.

5. Use the maximum standard deviation threshold algorithm to statistically analyze the point cloud clusters obtained by clustering. According to the characteristics of the actual size of the cone bucket, set the maximum standard deviation thresholds in the X, Y, and Z directions to filter out qualified point cloud clusters Think cone barrel.

Specifically, the maximum standard deviation threshold algorithm is used to statistically analyze the point cloud clusters obtained by clustering, and according to the characteristics of the actual size of the cone bucket, the maximum standard deviation thresholds in the X, Y, and Z directions are set, and the qualified points are screened out. The point cloud cluster of point cloud is regarded as a cone bucket, including the steps: according to the shape and size characteristics of the cone bucket, since the cone bucket can be regarded as a rotating body around the central axis in the vertical direction, the characteristics in the X and Y directions are consistent, and the maximum standard deviation The threshold is set to 0.08m, the height of the cone on the Z axis is 30cm, and its maximum standard deviation threshold is set to 0.15m, when the statistics of the X, Y, and Z directions of all points in the point cloud cluster obtained by clustering When the standard deviation of the coordinate values is less than the corresponding threshold, the point cloud cluster is considered as a cone bucket that meets the conditions.

6. Perform statistical analysis on the detected cone buckets, find one cone bucket on the left and right sides closest to the lidar, and obtain the coordinates of the center point of the point cloud cluster. Specifically, count the points of the cone buckets detected in the previous step, and calculate the average Y coordinate Mean_Y of all points in each cone bucket. Traverse all detected cone buckets, find the cone bucket with the largest number of points in the cone bucket whose average Y coordinate is greater than 0, that is, Mean_Y>0, and consider it to be the nearest cone bucket on the left side of the lidar, and discard the rest of the point cloud clusters. In the same way, find the cone with the most points in the cone with the average Y coordinate less than 0, that is, Mean_Y<0, and consider it to be the nearest cone on the right side of the lidar.

7. Calculate the average value of the coordinates of the center points of the left and right cones, and get the center point of the nearest left and right cones as the nearest target in the current state, and control the car to move towards the target. Specifically, calculate the average values of X, Y, and Z coordinates for the point cloud clusters of the nearest cone buckets in the left front and right front of the lidar obtained in the previous step, and use them as the spatial coordinate positions of the nearest cone buckets in the left front and right front, and calculate The coordinates of the center point of the connecting line are used as the moving target point of the robot in the current state.

8. Repeat the above steps to continuously obtain and track the latest target points. According to the above steps, the next moving target point in the current state of the vehicle is calculated, and the vehicle heading angle is adjusted according to its coordinate position relative to the vehicle, and the vehicle is controlled to move towards the target point. At the same time, the lidar acquires the next frame of point cloud data, starts a new point cloud data processing step, obtains the next moving target point, and controls the vehicle to move forward. In the ROS robot operating system, the above steps of point cloud processing and vehicle control are continuously cycled to realize the track cone detection and target point tracking based on multi-line lidar.

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto, any person familiar with the technical field within the scope disclosed in the present invention, according to the technical scheme of the present invention and its Any equivalent replacement or change of the inventive concept falls within the protection scope of the present invention.

Claims (10)

1. A racetrack cone detection and target point tracking method based on multi-line laser radar, characterized in that it comprises the following steps: 1) Read the lidar point cloud data under the ROS robot operating system; 2) Use a pass-through filter to perform pass-through filtering on the lidar point cloud data according to different track scenarios; 3) Adopt the random sampling consensus algorithm to eliminate the interference of the ground point cloud data on the cone bucket detection; 4) Use the Euclidean clustering extraction algorithm to initially screen out the point cloud clusters of the cone bucket; 5) Use the method of the maximum standard deviation threshold to perform statistical analysis on the point cloud clusters obtained by clustering. According to the characteristics of the actual size of the cone bucket, set the maximum standard deviation threshold in the X, Y, and Z directions to filter out qualified point clouds Clusters are considered cone buckets; 6) Statistically analyze the detected cone buckets, find one cone bucket on the left and right sides closest to the lidar, and obtain the coordinates of the center point of the point cloud cluster; 7) Calculate the average value of the center point coordinates of the cone barrels on the left and right sides of the laser radar, and obtain the center point of the nearest cone barrel on both sides of the laser radar as the nearest target point in the current state, and control the car to move towards the target point; 8) Cycle step 1) to step 7) to continuously acquire and track the latest target point. 2. A kind of racetrack cone detection and target point tracking method based on multi-line laser radar according to claim 1, characterized in that, step 1) specifically includes, on a computing platform equipped with an ubuntu16.04 operating system , install the ROS (Robot Operating System) robot operating system, and configure the lidar driver package, start the lidar running node for real-time collection of point cloud data, according to the characteristics of the track, the size of the cone bucket to be detected is a*a*b, In order to enable the lidar to scan on the cone to the greatest extent, the installation position of the lidar is located under the front nose of the racing car, and the installation height and ground clearance are b/2. 3. A method for cone detection and target point tracking based on multi-line laser radar according to claim 1, characterized in that: step 2) includes, before starting the detection, presetting the track scene to be detected , including 75-meter linear acceleration track, 8-figure circle track, and high-speed tracking track; set the detection range of lidar for different tracks, and filter out the point cloud data outside the range. 4. A method for cone detection and target point tracking based on multi-line laser radar according to claim 1, characterized in that: step 3) includes, using the random sampling consensus algorithm to set the plane filter threshold to a/5 , to judge the maximum deviation distance of the points in the normal direction of the point cloud cluster extracted by the algorithm, and consider the point cloud cluster with the maximum deviation distance greater than a/5 as not a plane, and the maximum deviation distance in the normal direction of the point cloud cluster will be not equal to The point cloud clusters within a range greater than a/5 are considered to be the same plane, and the point cloud clusters belonging to the plane are removed in the current point cloud to achieve the purpose of filtering the ground point cloud data. 5. A method for cone detection and target point tracking based on multi-line laser radar according to claim 1, characterized in that: step 4) is specifically to analyze the pre-collected point cloud data, and use laser radar to collect The number of points on different cones on the track, the nearest cone point is R, and the farthest can be scanned is r; therefore, according to the statistical law of several times of data collection, the minimum number of clustering points is set to r points, the maximum number of clustering points is R points, set the maximum distance between two points to be searched in the clustering process as L, use KD tree for point cloud search, and the clustering condition will be met, that is, the number of points is within the range of (r, R) The point cloud of the point cloud is divided into several point cloud clusters; the number of points in each point cloud cluster is counted, and the average value of X, Y, and Z coordinates is calculated for each point cloud cluster as the point cloud center of gravity of the point cloud cluster , which is used to replace the position of the point cloud cluster relative to the lidar. 6. A method for detecting cones and target points on a track based on multi-line lidar according to claim 1, characterized in that: Step 5) specifically includes, according to the shape and size characteristics of the cones, since the cones are The rotating body around the central axis in the vertical direction has a width of a in the X and Y directions, so the maximum standard deviation threshold in the X and Y directions is set to Q, the height of the cone on the Z axis is b, and the Z axis is set to The maximum standard deviation threshold of q is set to q, when the standard deviation of the X, Y, and Z coordinate values of all points in the clustered point cloud cluster is less than the corresponding threshold, that is, when the calculated point cloud cluster When the standard deviation in the X and Y directions of is less than Q and the standard deviation in the Z direction is less than q, the point cloud cluster is considered to be a cone bucket that meets the conditions. 7. A method for cone detection and target point tracking based on multi-line laser radar according to claim 1, characterized in that: step 6) specifically includes counting the number of cones detected in step 5) , and calculate the average Y coordinate of all points in each cone bucket; traverse all detected cone buckets, find the cone bucket with the largest number of points in the cone bucket whose average Y coordinate is greater than 0, and consider it to be the nearest cone bucket on the left side of the lidar. The reason is to find the cone with the most points in the cone with the average Y coordinate less than 0, which is considered to be the nearest cone on the right side of the lidar. 8. A method for cone detection and target point tracking based on multi-line laser radar according to claim 1, characterized in that: Step 7) specifically includes calculating the cones closest to the left front and right front of the laser radar respectively The average value of the X, Y, and Z coordinates of the point cloud cluster is considered as the center point coordinates of the nearest left and right cone barrels of the lidar, and the midpoint coordinates of the connecting line between the left and right cone barrels of the lidar are calculated, and it is used as the target in the current state Points, the specific steps include, for the point cloud clusters of the nearest cone buckets in the left front and right front of the lidar obtained in step 6), respectively calculate the average values of X, Y, and Z coordinates, as the space of the nearest cone buckets in the left front and right front Coordinate position, and calculate the coordinates of the center point of its connection, as the movement target point of the robot in the current state. 9. A method for cone detection and target point tracking based on multi-line laser radar according to claim 1, characterized in that: step 8) specifically includes performing a test on each frame of laser radar under the ROS robot operating system The collected point cloud data is processed from step 1) to step 7), and after the data processing of each frame is completed, the moving target point of the next frame of the robot is output, and the robot is controlled to move towards the target point. At the same time, the next frame The data collected by the lidar is also processed from step 1) to step 7) to achieve the purpose of real-time detection of cones and tracking of target points. 10. The method for cone detection and target point tracking based on multi-line laser radar according to claim 1, characterized in that: the laser radar is a 16-line laser radar.

Images