CN117890882A - Method and system for identifying pedestrians by 2D laser radar point cloud - Google Patents
Method and system for identifying pedestrians by 2D laser radar point cloud Download PDFInfo
- Publication number
- CN117890882A CN117890882A CN202410270594.1A CN202410270594A CN117890882A CN 117890882 A CN117890882 A CN 117890882A CN 202410270594 A CN202410270594 A CN 202410270594A CN 117890882 A CN117890882 A CN 117890882A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- radar
- distance
- candidate
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000012216 screening Methods 0.000 claims abstract description 44
- 238000004422 calculation algorithm Methods 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 abstract description 9
- 238000004364 calculation method Methods 0.000 abstract description 6
- 230000002093 peripheral effect Effects 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 18
- 230000006870 function Effects 0.000 description 12
- 238000001514 detection method Methods 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 4
- 238000005259 measurement Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 230000004888 barrier function Effects 0.000 description 2
- 238000005452 bending Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 239000000523 sample Substances 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 235000019580 granularity Nutrition 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- LWYJUZBXGAFFLP-OCNCTQISSA-N menogaril Chemical compound O1[C@@]2(C)[C@H](O)[C@@H](N(C)C)[C@H](O)[C@@H]1OC1=C3C(=O)C(C=C4C[C@@](C)(O)C[C@H](C4=C4O)OC)=C4C(=O)C3=C(O)C=C12 LWYJUZBXGAFFLP-OCNCTQISSA-N 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
Abstract
The invention provides a method and a system for identifying pedestrians by using a 2D laser radar point cloud, wherein the method comprises the following steps: acquiring 2D laser radar data; clustering the 2D laser radar data to obtain radar point cloud clusters; screening out a first candidate point cloud cluster according to the curvature of the radar point cloud cluster and the distance from the radar center; screening out a second candidate point cloud cluster which has the curvature with the first candidate point cloud cluster and the distance from the radar center meeting a threshold and is closest to the first candidate point cloud cluster; the first subsequent point cloud and the second candidate point cloud form a pedestrian. According to the method, the 3D radar point cloud is not required, the requirement on the peripheral performance of the robot is low, the equipment cost can be greatly saved, the laser radar point cloud identification is not required to be performed in a deep neural network mode or the like, the calculation force can be saved, pedestrians can be accurately identified on a hardware platform with limited cost, and the method has strong competitiveness.
Description
Technical Field
The invention relates to the technical field of radar point cloud data processing, in particular to a method and a system for identifying pedestrians by using 2D laser radar point clouds.
Background
LiDAR (LiDAR) includes a laser ranging system, an optical mechanical scanning unit, a control recording unit, a global positioning system (Global Position System, GPS), an inertial measurement system (Inertial Measurement Unit, IMU), a set of imaging devices, and the like. Laser radar uses laser energy to measure the distance between the sensor and the object being measured. Most lidar systems are pulse-based. The basic principle of pulsing is based on measuring the time it takes to emit a pulse of laser energy and then to reach the target. Under the fire and heat of robots and unmanned, the laser radar is widely applied. The 2D laser radar is divided into two types of trigonometry and TOF according to the principle, and generally consists of a laser transmitter, a receiver, a signal processing unit and a motor. The laser transmitter is responsible for transmitting laser, the wavelength is 905nm as a main part, and 850nm is adopted for a manufacturer at a short distance. The receiver is responsible for receiving light reflected back from the obstruction. The signal processing unit is responsible for controlling the emission of the laser and processing the return signal, and calculates the distance of the target object. The motor is responsible for the rotation of the laser, and the scanning of 360 degrees on the same plane is realized, so that a two-dimensional map is established. The basic principle of the triangular laser radar is as follows: the laser emits laser to the measured object, the laser is converged on the CCD sensor through the lens after being reflected by the object, when the measured object moves, the light spot on the sensor also moves, and the distance from the laser to the object can be obtained through the geometric triangle relation. Compared with a triangular radar, the TOF radar principle is simpler, a laser emits laser signals to an object to be measured, the laser signals reach a photoelectric device after reflection, the signals are converted and are transmitted to an echo signal processing circuit, finally electric pulse signals are obtained and transmitted to a timer, the time of light flight is calculated, and the distance from the radar to the object to be measured is calculated by dividing the total time from the departure of measuring light to the reception of measuring light by 2 times the light speed because the light speed is fixed.
The robot needs to identify whether pedestrians exist in the environment in the running process, and the obstacle avoidance strategy is adjusted in real time under the condition that pedestrians exist, so that the robot can be better prevented from rubbing the pedestrians, and the running safety and experience of the robot are improved. In the prior art, a conventional identification strategy is to identify laser radar point clouds or images by adopting a camera and a three-dimensional radar through a deep neural network and other modes. The prior art has the defects that the cost of the 3D laser radar detection equipment adopted by the robot is too high, or the resource consumption of the GPU and the like which need a host computer is also high through a neural network, and the method is not suitable for a hardware platform scheme with limited cost.
Disclosure of Invention
The invention aims to provide a method and a system for identifying pedestrians by using 2D laser radar point clouds, wherein the method adopts the 2D laser radar to identify pedestrians in the moving process of a robot, clusters 2D laser radar data to form the laser radar point clouds by receiving the 2D laser radar data, and judges whether each point cloud is a pedestrian point cloud according to the curvature of the point clouds and the distance between the point clouds and a radar center.
A method for identifying pedestrians by a 2D lidar point cloud, comprising:
acquiring 2D laser radar data;
clustering the 2D laser radar data to obtain radar point cloud clusters;
screening out a first candidate point cloud cluster according to the curvature of the radar point cloud cluster and the distance from the radar center;
screening out a second candidate point cloud cluster which has the curvature with the first candidate point cloud cluster and the distance from the radar center meeting a threshold and is closest to the first candidate point cloud cluster;
the first candidate point cloud cluster and the second candidate point cloud cluster form a pedestrian.
Preferably, clustering the 2D lidar data to obtain a cloud of radar points includes:
sequentially comparing the distances of two continuous radar points according to the index sequence of the radar data, and when the distance between the two radar points exceeds a threshold value, considering that the two radar points do not belong to the same radar point cloud cluster;
after the radar point cloud is formed, judging whether the number of points in the radar point cloud meets the screening requirement, and when the number of points in the radar point cloud is smaller than or larger than a threshold value, eliminating the radar point cloud which does not meet the requirement.
Preferably, the screening the first candidate point cloud according to the curvature of the radar point cloud and the distance from the radar center comprises:
Calculating curvature of the radar point cloud: inputting coordinates of a starting point, a center point and an end point of the radar point cloud;
calculating the distance difference between the starting point abscissa and the intermediate point abscissa;
calculating the distance difference between the ordinate of the starting point and the ordinate of the intermediate point;
calculating the distance difference between the abscissa of the ending point and the abscissa of the intermediate point;
calculating the distance difference between the ordinate of the end point and the ordinate of the middle point;
calculating a first arctangent value of a distance difference between the starting point ordinate and the intermediate point ordinate divided by a distance difference between the starting point abscissa and the intermediate point abscissa;
calculating a second arctangent value of the difference between the distances of the ending point ordinate and the intermediate point ordinate divided by the difference between the distances of the ending point abscissa and the intermediate point abscissa;
the curvature value is the difference between the first and second arctangent values plus 3.14, multiplied by 6.28%, and finally subtracted by the absolute value of 3.14.
Preferably, the screening the first candidate point cloud according to the curvature of the radar point cloud and the distance from the radar center comprises:
calculating the distance between the radar point cloud and the radar center, and when the distance between the radar point cloud and the radar center is larger than a preset distance, conforming the current radar point cloud to the screening requirement of the first candidate point cloud;
And when one radar point cloud group meets the curvature screening requirement and the distance screening requirement simultaneously, marking the current radar point cloud group as a first candidate point cloud group.
Preferably, screening out a second candidate point cloud that meets a threshold value with respect to curvature of the first candidate point cloud and a distance from a radar center and is closest to the first candidate point cloud includes:
after the first candidate point cloud cluster is selected, continuing to screen the radar point cloud clusters which are not screened;
judging the distance between the current radar point cloud and the first candidate point cloud when the difference between the curvature of the radar point cloud and the curvature of the first candidate point cloud and the difference between the distance between the radar point cloud and the radar center of the first candidate point cloud and the distance between the radar point cloud and the radar center of the first candidate point cloud are smaller than a threshold value;
and when the distance between the current radar point cloud and the first candidate point cloud is smaller than a threshold value, marking the current radar point cloud as a second candidate point cloud.
Preferably, the distances of two consecutive radar points are compared in turn in the index order of the radar data, and when the distance between two radar points exceeds a threshold value, the two radar points are not considered to belong to the same cloud of radar points, including:
acquiring coordinates of a radar point P1 and a radar point P2;
Calculating a distance d between coordinates of the radar point P1 and the radar point P2;
when d is greater than 0.32, two radar points do not belong to the same radar point cloud, and when d is less than 0.32, two radar points belong to the same radar point cloud.
Preferably, after the 2D laser radar data is acquired, noise reduction is performed on the 2D laser radar data, specifically:
obtaining parameters of Eps and MinPts in a DBSCAN algorithm;
establishing Kd-tree acceleration neighborhood search, and performing DBSCAN clustering denoising on 2D laser radar data noise;
setting the order of a moving least square method, carrying out normal estimation, and determining a fitting radius;
smoothing and denoising the 2D laser radar data by adopting a mobile least square method;
and outputting the noise-reduced 2D laser radar data.
A system for 2D lidar point cloud identification of pedestrians, comprising:
the data acquisition module is used for acquiring 2D laser radar data;
the data processing module is used for clustering the 2D laser radar data to obtain radar point cloud clusters;
the first judging module is used for screening out first candidate point cloud clusters according to the curvature of the radar point cloud clusters and the distance from the radar center;
the second judging module is used for screening out a second candidate point cloud cluster which meets a threshold value with the curvature of the first candidate point cloud cluster and the distance from the center of the radar and is closest to the first candidate point cloud cluster;
And the data output module is used for forming a pedestrian by the first candidate point cloud cluster and the second candidate point cloud cluster.
An electronic device, comprising: the system comprises a chip, a processor and a memory, wherein the memory is used for storing computer program codes, the computer program codes comprise computer instructions, and the electronic equipment executes a method for identifying pedestrians by using 2D laser radar point clouds under the condition that the chip executes the computer instructions.
A computer readable storage medium having stored therein a computer program comprising program instructions which, when executed by a processor of an electronic device, cause the processor to perform a method of 2D lidar point cloud recognition of pedestrians.
According to the invention, the 2D laser radar is adopted to externally acquire the point cloud for identifying pedestrians in the running process of the robot, the 2D laser radar can acquire the two-dimensional point cloud image, namely, the point cloud image is formed on the walking plane of the robot, and because the path planning is required to be carried out in real time based on the situation of pedestrians on the road surface in the walking process of the robot, the 2D laser radar can meet the requirement of pedestrian obstacle avoidance, the 3D laser radar is not required to be adopted for detection, a large amount of equipment throwing cost can be saved, and the 2D laser radar is convenient to calculate and process due to the small data quantity, in the prior art, the depth neural network is adopted to screen the pedestrians, so that the calculation force burden is seriously increased, the 2D laser radar can accurately calculate the positions of the pedestrians through the transverse and longitudinal coordinates of the point cloud of each laser radar, and the robot is controlled to move in real time according to the positions of the pedestrians, so that the pedestrian obstacle avoidance is completed.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, and it will be obvious to a person skilled in the art that other drawings can be obtained from these drawings without inventive effort.
FIG. 1 is a schematic flow chart of the present invention;
FIG. 2 is a schematic diagram of a 2D lidar data clustering process of the present invention;
FIG. 3 is a schematic diagram of a second candidate point cloud screening process according to the present invention;
FIG. 4 is a schematic diagram of a distance clustering process of a radar point cloud image according to the present invention;
fig. 5 is a schematic diagram of a denoising process of 2D lidar data according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that all directional indicators (such as up, down, left, right, front, and rear … …) in the embodiments of the present invention are merely used to explain the relative positional relationship, movement, etc. between the components in a particular posture (as shown in the drawings), and if the particular posture is changed, the directional indicator is changed accordingly.
Furthermore, the description of "first," "second," etc. in this disclosure is for descriptive purposes only and is not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In addition, the technical solutions of the embodiments may be combined with each other, but it is necessary to base that the technical solutions can be realized by those skilled in the art, and when the technical solutions are contradictory or cannot be realized, the combination of the technical solutions should be considered to be absent and not within the scope of protection claimed in the present invention.
The robot needs to identify whether pedestrians exist in the environment in the running process, and the obstacle avoidance strategy is adjusted in real time under the condition that pedestrians exist, so that the robot can be better prevented from rubbing the pedestrians, and the running safety and experience of the robot are improved. In the prior art, a conventional identification strategy is to identify laser radar point clouds or images by adopting a camera and a three-dimensional radar through a deep neural network and other modes. The prior art has the defects that the cost of the 3D laser radar detection equipment adopted by the robot is too high, or the resource consumption of the GPU and the like which need a host computer is also high through a neural network, and the method is not suitable for a hardware platform scheme with limited cost.
According to the invention, the 2D laser radar is adopted to externally acquire the point cloud for identifying pedestrians in the running process of the robot, the 2D laser radar can acquire the two-dimensional point cloud image, namely, the point cloud image is formed on the walking plane of the robot, and because the path planning is required to be carried out in real time based on the situation of pedestrians on the road surface in the walking process of the robot, the 2D laser radar can meet the requirement of pedestrian obstacle avoidance, the 3D laser radar is not required to be adopted for detection, a large amount of equipment throwing cost can be saved, and the 2D laser radar is convenient to calculate and process due to the small data quantity, in the prior art, the depth neural network is adopted to screen the pedestrians, so that the calculation force burden is seriously increased, the 2D laser radar can accurately calculate the positions of the pedestrians through the transverse and longitudinal coordinates of the point cloud of each laser radar, and the robot is controlled to move in real time according to the positions of the pedestrians, so that the pedestrian obstacle avoidance is completed.
Example 1
A method for identifying pedestrians by a 2D lidar point cloud, comprising:
s100, acquiring 2D laser radar data;
The 2D laser radar is also called a single-line radar, the 2D laser radar is generally carried on a warehousing and catering service robot and a sweeping robot, the 2D laser radar has a wide application range, and the 2D laser radar is generally structured by a pair of laser transmitting and receiving heads and a motor with an encoder, wherein the motor carries the sensor. When in operation, the motor rotates at a constant speed, the laser transmitter transmits a probe beam at a fixed frequency, and the receiving head records the echo of the device at a specific angle to calculate the distance.
S200, clustering the 2D laser radar data to obtain radar point cloud clusters;
unlike image data, radar data is sparse, and returned data only has position information of a reflection point from the laser radar. The data obtained by the 2D laser radar are all regarded as being on a plane, so the application adopts an image processing method to process the obtained 2D laser radar points, the grid/pixel with the obstacle (the laser reflection point position) in the robot working scene is set as 1, and the pixels at the laser path points are all 0, so that a binary image is obtained. When the 2D laser radar detects an obstacle, the returned points can be densely distributed at a certain position, and a plurality of 2D laser radar data points represent the position of the obstacle; when the 2D laser radar does not detect the obstacle, the returned points are uniformly distributed, so that whether the obstacle exists or not can be judged according to the distribution condition of the returned points of the 2D laser radar, 2D laser radar data are required to be clustered, and 2D laser radar data points of the same obstacle are classified into one type to obtain radar point cloud clusters.
In this application, clustering methods for 2D lidar data points include, but are not limited to: k-means, the principle of which is to find the mass center (center point) of the data point and distribute the data point to the nearest mass center, so as to continuously update the position of the mass center to achieve the best clustering effect, wherein the data points are grouped into K clusters so that the points in the clusters are similar as much as possible and the points among the clusters are different as much as possible. Hierarchical clustering, which is performed by constructing a hierarchical structure between data points, can be divided into a bottom-up aggregation method and a top-down splitting method, and can process data point division of different levels and granularities. Density-based clustering, such as DBSCAN, can identify clusters of arbitrary shape and has better robustness to noise and outliers. One method divides the data by computing clusters in which data points of high density are located. Spectral clustering, which uses a similarity matrix of data to perform clustering, is particularly suitable for a data set with a complex shape, and finds the similarity between data points by calculating characteristic values and characteristic vectors. The Gaussian mixture model is a clustering method based on a probability model, is suitable for estimating the distribution of sub-groups, and can be used for simulating the distribution of potential variables in a multidimensional dataset. Fuzzy C-means, similar to K-means, but allows a data point to belong to multiple clusters, each cluster having a certain degree of membership or probability. K-means, similar to K-means, uses data points instead of means as the center of the cluster. Mean Shift, the region with highest data point density is found by iteratively updating candidate cluster center points.
S300, screening out first candidate point cloud clusters according to the curvature of the radar point cloud clusters and the distance from the radar center;
the 2D laser radar detects different obstacles, radar point clouds in different forms can be reflected according to parameters (shape, size and distance) of the obstacles, and data obtained by scanning of the 2D laser radar system comprise: the method comprises the steps of obtaining the surface area of a target expressed by radar point cloud through calculation of data obtained through scanning of a 2D laser radar system, wherein the principle of identifying pedestrians is mainly achieved through scanning the feet of the pedestrians, the feet of the pedestrians are similar to cylinders unlike a plane like a wall surface, radar point cloud formed by the feet of the pedestrians is of a certain curvature, the point cloud of the pedestrians is only possible if the points of the certain curvature exist, the distance between the radar point cloud and the radar center is within a certain range, the measuring accuracy is not high because the distance between the radar point cloud and the laser radar is too far, obstacle avoidance of the robot is not affected, the too far point cloud is not considered, the distance between the pedestrians and the radar center is more than 0.1 meter, and in order to prevent signal interference, the coordinates of the radar center are set to be (0, 0).
S400, screening out a second candidate point cloud cluster which has the curvature with the first candidate point cloud cluster and the distance from the radar center meeting a threshold and is closest to the first candidate point cloud cluster;
since a person has two feet, each foot forms a radar point cloud, one foot of the same person can be determined through the screening of the previous step, the second foot of the natural person needs to be screened out because the two feet can be used for determining the person, the parameters of the radar point cloud formed by the feet of each person are different because the feet of each person are different, the size and the length of each foot of each person can influence the area and the curvature of the formed radar point cloud, but the area and the curvature of the radar point cloud formed by the two feet of the one person are similar, and the distance between the radar point clouds formed by the two feet of the one person is smaller, so that the radar point cloud similar to the radar point cloud of the foot which is determined as the pedestrian is screened out in a certain distance range, and the radar point cloud similar to the radar point cloud of the foot which is determined as the pedestrian is used as the two feet of the current person, thereby determining a complete person.
S500, the first candidate point cloud cluster and the second candidate point cloud cluster form a pedestrian.
The first candidate point cloud cluster and the second candidate point cloud cluster are combined to form a point cloud cluster serving as a pedestrian point cloud cluster, if the point cloud cluster is single, a path planned by the robot possibly passes through the middle of two feet, and obviously, the situation is not feasible in an actual application scene, so that the point cloud clusters formed by the left foot and the right foot of a pedestrian are combined to perform obstacle avoidance judgment, and the robot can accurately avoid the pedestrian during normal operation.
Preferably, S200, clustering the 2D laser radar data to obtain a radar point cloud includes:
s210, sequentially comparing distances of two continuous radar points according to an index sequence of radar data, and when the distance between the two radar points exceeds a threshold value, considering that the two radar points do not belong to the same radar point cloud;
radar data (lidar data) is an array of 2*n, with the first row of data being x-axis data and the second row of data being y-axis data. The center coordinates of the radar body are (0, 0), and when the robot runs at a working place, the 2D laser radar detection device arranged above the front part of the robot is pressed into environment data obtained by radar scanning from-180 degrees to 180 degrees in sequence. Since the foot size of a pedestrian is within a certain range, the distribution of radar points returned by the foot of the pedestrian is also densely distributed within a certain range, so that if the distance between two radar points exceeds a threshold value, the two radar points are not considered to belong to the same cloud of radar points, i.e. if the distance between two radar points exceeds the threshold value, the two radar points are not considered to belong to the same foot.
The point cloud includes: sparse point clouds, multi-source point clouds, and dense point clouds, sparse point clouds: early stage is limited by matching technology and computing power, and the photogrammetric point cloud is sparse and small in size. At that time, the types of laser scanning systems are limited, and have not been widely used. The main stream laser data ALS point cloud is also sparse. The point cloud at this stage cannot represent the surface coverage of the object level, limited by the point density. Thus, there is no specific need for a sophisticated PCS or PCSS. Researchers have focused mainly on three-dimensional mapping (DEM generation) and simple object extraction (e.g., rooftop).
Dense point clouds: and the computer vision algorithms such as dense matching and the like, and the efficient point cloud generators such as various LiDAR systems, RGB-D sensors and the like start dense point clouds. Dense and large-volume point clouds create more possibilities in 3D applications, but there is also a need for stronger practical algorithms.
Multiple source point clouds: new generation point clouds, such as satellite photogrammetry point clouds and TomoSAR point clouds, stimulate the need for correlation algorithms. Different point clouds have different characteristics and application environments.
Lidar point clouds are the most commonly used data in PCS. They have been widely used for buildings (urban environments) and trees (forests), which are also the most popular research targets in traditional PCs.
The 2D navigation laser radar can only navigate in real time basically, and two-dimensional imaging lacks height information and cannot image. The 3D laser radar can perform three-dimensional dynamic real-time imaging, can perform real-time imaging, and can restore the shape and the size of an object and restore the three-dimensional information of space. So that the automatic driving is used, and the three-dimensional dynamic laser radar is used. However, the 3D laser radar has more acquired data, high calculation force requirement and high algorithm requirement, and the release cost of the 3D laser radar is at least higher than that of the 2D laser radar by a plurality of times. Is not suitable for low-cost development environments. Although the 2D laser radar lacks of height information and cannot image, the 2D laser radar can detect ground obstacles in real time, and the requirement of real-time navigation is met.
S220, after the radar point cloud is formed, judging whether the number of points in the radar point cloud meets the screening requirement, and when the number of points in the radar point cloud is smaller than or larger than a threshold value, eliminating the radar point cloud which does not meet the requirement.
After a plurality of radar points are clustered into one radar point cloud, judging whether the number of radar points in the formed radar point cloud meets the requirement, if the number of radar points in the radar point cloud is too large or too small, considering that the current radar point cloud is not a radar point cloud formed by pedestrians, finally, screening the number of radar points in the radar point cloud to be a preliminary pedestrian radar point cloud, and then, carrying out secondary screening to be a formal pedestrian radar point cloud, wherein the robot carries out obstacle avoidance path planning according to the formal pedestrian radar point cloud selected after the secondary screening, in the application, a simpler algorithm is adopted as a radar point preliminary processing algorithm, namely a distance algorithm and a number algorithm, firstly, the distance between single radar points is calculated to be the radar point cloud, then, a counting algorithm is adopted, the number of radar points in the partitioned radar point cloud is calculated, and whether the radar point cloud is the number of the pedestrian radar point cloud can be the formal pedestrian radar point cloud, the robot can carry out obstacle avoidance path planning according to the formal pedestrian radar point cloud selected after the secondary screening, in the application, the distance algorithm is carried by one algorithm and the two algorithm is not required to be carried in the cloud, and the application has a great cost is greatly reduced, and the distance between the two algorithm is carried in the application is greatly required, and the application is 200.
Preferably, S300, screening the first candidate point cloud according to the curvature of the radar point cloud and the distance from the radar center includes:
calculating curvature of the radar point cloud: inputting coordinates of a radar point cloud starting point A, a center point B and an end point C; x is X A X is the abscissa of the starting point A B X is the abscissa of the center point B C Is the abscissa of the end point C, Y A Is the ordinate of the starting point A, Y B Is the ordinate of the center point B, Y C Is the ordinate of the end point C.
Calculating a distance difference D1 between the starting point abscissa and the intermediate point abscissa;
D1=X B -X A
calculating a distance difference D2 between the ordinate of the starting point and the ordinate of the intermediate point;
D2=Y B -Y A
calculating a distance difference D3 between the abscissa of the end point and the abscissa of the middle point;
D3=X C -X A
calculating a distance difference D4 between the ordinate of the end point and the ordinate of the intermediate point;
D4=Y C -Y A
calculating a first arctangent value K1 of the distance difference between the starting point ordinate and the intermediate point ordinate divided by the distance difference between the starting point abscissa and the intermediate point abscissa;
k1 =arctan (D2/D1, arctan is an arctangent function
Calculating a second arctangent value K2 of the difference between the end point ordinate and the intermediate point ordinate divided by the difference between the end point abscissa and the intermediate point abscissa
K2arctan（D4/D3）
The curvature K is the difference between the first and second arctangent plus 3.14, multiplied by 6.28%, and finally subtracted by 3.14 absolute.
K=|（（K1-K2）+3.14）6.28%-3.14|
The curvature of a curve is defined by differentiation for the rotation rate of the tangential angle of a certain point on the curve with respect to the arc length, and indicates the degree of deviation of the curve from a straight line, and the larger the numerical curvature which indicates the bending degree of the curve at the certain point, the larger the bending degree of the curve is. The inverse of the curvature is the radius of curvature. The radius of curvature of an arc is the radius of a circle formed when the arc is a part of the circle. The larger the radius of curvature, the flatter the arc, the smaller the radius of curvature, and the steeper the arc. The inverse of the radius of curvature is the curvature. Curvature is defined as (angle rotated/corresponding arc length). When both the angle and the arc length approach 0, it is a standard definition of the curvature of a smooth curve of arbitrary shape. Whereas for a circle the curvature does not vary with position. The radar point cloud formed by the feet of the pedestrians, whether scanned from the front or from the back, is arc-shaped, which is the biggest feature of the pedestrians in distinguishing the point cloud from other obstacle point clouds, such as: the wall is a straight line, the curvature of which is obviously much smaller than the curvature of the foot of a pedestrian; the table leg is a cylinder shape which is far smaller than the foot of the pedestrian, the curvature of the table leg is obviously far greater than the curvature of the foot of the pedestrian, and although the foot of each pedestrian presents different arcs, the difference of the foot of each person is within 10cm, and the difference of the curvature is not great, so that the curvature value of the radar point cloud cluster in a certain range is judged as the foot of the pedestrian,
Preferably, S300, screening the first candidate point cloud according to the curvature of the radar point cloud and the distance from the radar center includes:
calculating the distance between the radar point cloud and the radar center, and when the distance between the radar point cloud and the radar center is larger than a preset distance, conforming the current radar point cloud to the screening requirement of the first candidate point cloud;
after calculating the curvature of the radar point cloud, the distance between the radar point cloud and the radar center is also calculated, and as a preferred embodiment, the present invention selects the center point B of the radar point cloud as the coordinates of the whole radar point cloud, and then calculates the distance between the coordinates of the radar point cloud and the radar center (0, 0).
And when one radar point cloud group meets the curvature screening requirement and the distance screening requirement simultaneously, marking the current radar point cloud group as a first candidate point cloud group.
When the distance between the coordinates of the radar point cloud and the radar center (0, 0) is greater than 0.1 m and less than 5 m, the current radar point cloud is marked as a first candidate point cloud, and because the coordinates of the radar point cloud are far away from the radar, the measurement accuracy is not high, and obstacle avoidance of a robot is not affected, so that the far-away radar point cloud is not considered, in the embodiment of the invention, the radar point cloud exceeding 5 m is the radar point cloud far away from the radar center point, the radar point cloud is excluded from the first candidate point cloud, and if the coordinates of the radar point cloud are far away from the radar center point by less than 0.1 m, signal interference is caused, and in the embodiment of the invention, the radar point cloud is excluded from the first candidate point cloud, and the radar point cloud which meets the curvature requirement is used as the first candidate point cloud except the radar point cloud far away from the radar center point cloud (0, 0).
Preferably, S400, screening out a second candidate point cloud that meets the threshold value with respect to the curvature of the first candidate point cloud and the distance from the radar center and is closest to the first candidate point cloud includes:
s410, after the first candidate point cloud cluster is selected, continuing to screen the radar point cloud clusters which are not screened;
after the first candidate point cloud cluster (one foot of the pedestrian) is screened, the other foot of the pedestrian (the second candidate point cloud cluster) needs to be found, so that the parameters of the next radar point cloud cluster are calculated, and whether the parameters are the second foot of the same person is judged.
S420, judging the distance between the current radar point cloud and the first candidate point cloud when the difference between the curvature of the radar point cloud and the curvature of the first candidate point cloud and the difference between the distance between the radar point cloud and the radar center of the first candidate point cloud and the distance between the radar point cloud and the radar center of the first candidate point cloud are smaller than a threshold value;
the two feet of a person are similar, so that the radar point cloud formed by the two feet of the person is similar, the similarity of the current radar point cloud and the first candidate point cloud can be judged according to the difference between the curvature of the radar point cloud and the curvature of the first candidate point cloud and the difference between the distance of the radar point cloud and the radar center of the first candidate point cloud and the difference between the curvature of the radar point cloud and the distance of the radar point cloud and the radar center of the first candidate point cloud is smaller than a threshold value, and when the difference between the curvature of the radar point cloud and the curvature of the first candidate point cloud and the distance of the radar point cloud from the radar center of the first candidate point cloud is larger than the threshold value, the difference between the distance of the radar point cloud and the radar center of the first candidate point cloud is not similar to the first candidate point cloud.
When the similarity between the current radar point cloud and the first radar point cloud is high, judging the distance between the current radar point cloud and the first radar point cloud is started, and in an actual scene, whether the two radar point clouds are the two feet of the same person or not can be judged by calculating the distance between the current radar point cloud and the first radar point cloud because the distance between the two feet of the same person is within a certain range.
And S430, when the distance between the current radar point cloud and the first candidate point cloud is smaller than a threshold value, marking the current radar point cloud as a second candidate point cloud.
When the distance between the current radar point cloud cluster and the first candidate point cloud cluster is smaller than a threshold value, the fact that the current radar point cloud cluster is the same second foot is indicated, the current radar point cloud cluster is marked as the second candidate point cloud cluster, two feet of a pedestrian are found, the obstacle influence range of the pedestrian can be planned through the two feet of the pedestrian, and the subsequent path planning algorithm is guided to conduct real-time obstacle avoidance of the robot.
Preferably, S210, in order of the index of the radar data, compares the distances of two consecutive radar points in turn, and when the distance between two radar points exceeds a threshold value, considers that the two radar points do not belong to the same radar point cloud includes:
S211, acquiring coordinates of a radar point P1 and a radar point P2;
s212, calculating a distance d between coordinates of the radar point P1 and the radar point P2;
s213, when d is larger than 0.32, the two radar points do not belong to the same radar point cloud, and when d is smaller than 0.32, the two radar points belong to the same radar point cloud.
Since the length of an adult foot is generally 25-32cm, when the distance d between the coordinates of the radar point P1 and the radar point P2 is greater than 0.32 meters, it means that the two radar points are not the same foot, the two radar points are divided into different feet, when the distance d between the coordinates of the radar point P1 and the radar point P2 is less than 0.32 meters, it means that the two radar points are the same foot, the two radar points are divided into one foot, and all the radar points belonging to one foot are found according to the distance d between the coordinates of the radar point P1 and the radar point P2, so as to jointly form a radar point cloud.
Preferably, S100, after acquiring the 2D laser radar data, further includes S110, noise reduction is performed on the 2D laser radar data, specifically:
s111, obtaining parameters of Eps and MinPts in a DBSCAN algorithm;
if most of the K most similar data sets in the feature space belong to a certain class, the sample also belongs to the class, namely noise and data belong to two classes, and the noise can be separated by the difference of the noise and the data in the feature space; calculating the characteristics of the unknown data and the characteristics corresponding to each piece of data in the data set D by adopting an Euclidean distance formula to obtain a distance distribution matrix of the data set D; then, the distance distribution matrixes are arranged in an ascending order to obtain a KNN distribution matrix, and average distances with average values of K values are obtained according to columns to obtain an Eps parameter list; the above Euclidean distance formula is defined as: Wherein D represents the distance between the 2D laser radar point and the feature, and x i Representing the characteristics which the 2D laser radar point should possess, y i Representing the corresponding characteristics of the 2D laser radar points;
the distance distribution matrix of the 2D lidar dataset D is defined as:
Dist n×n ＝{dist(i,j)|1≤i≤n,1≤j≤n}
wherein Dist n×n N is a real symmetric matrix of n×n, n is the number of objects in the 2D lidar dataset D; dist (dist) (i,j) Distance from the ith object to the jth object in the 2D laser radar data set D;
giving a 2D laser radar data set D, solving expected values of the number of data points in the neighborhood of the core point by taking Eps as the neighborhood, and generating MinPts parameters by adopting a mathematical expected method; the mathematical expectation method is defined as:
where n is the total number of core objects in the 2D lidar dataset D and Pi is the number of core objects in the Eps neighborhood of the ith object.
S112, establishing Kd-tree acceleration neighborhood search, and performing DBSCAN cluster denoising on 2D laser radar data noise;
substituting the threshold parameter into a DBACAN clustering algorithm, reading a 2D laser radar point cloud data set D, traversing the core points by taking Eps as a neighborhood, searching data points with density reaching the core points, clustering into clusters, and taking the Eps and MinPts parameters corresponding to the K value as optimal parameters when the number of clusters is stable and the density threshold is minimum; and (3) establishing Kd & # x2011, accelerating the tree to search for a neighborhood, reserving a cluster generated by a DBSCAN algorithm under the optimal parameters, deleting boundary points and noise points, and obtaining 2D laser radar point cloud data after removing large-scale noise. The 2D laser radar data points are constructed in a K-dimensional space, each layer in the tree corresponds to one dimension, kd & # x2011, and the tree is circularly constructed into a K-dimensional binary search tree through the attribute number of the data points. Kd & # x2011, the left subtree in the tree has a smaller value in a given dimension than the parent node, and the right subtree is larger than the parent node.
S113, setting a moving least square method order, performing normal estimation, and determining a fitting radius;
the specific mode for acquiring the specific parameters of the mobile least square method is as follows: establishing a fitting function of K neighbor neighborhood and local area, and setting the order of a basis function in the fitting function as 2; then determining the weight function as a cubic spline weight function, and fitting a moving least square curved surface to the neighborhood of each data point; calculating a curved surface in a local K neighborhood range, and carrying out normal estimation according to the curved surface; creating Kd & # x2011, searching a neighborhood by the tree, and determining the fitted K neighbor radius; and (5) reconstructing the curved surface to remove small-scale noise.
On one local sub-field of the fitting region, the fitting function f (x) is defined as:
wherein,＝[p 1 (x)p 2 (x)…p m (x)]as a basis function +.>(i=0, 1,) n is the coefficient vector to be solved, and the spatial coordinates corresponding to the coefficients are x= [ x, y, z]Function of->A vector of core objects that are Eps neighbors of the ith object.
The cubic spline weight function is defined as:
wherein r=x‑x i ，h i The range of the i-th node weight function influence area is that beta is an influence coefficient, and the value range is usually 1.5-2.5, and is usually 2.0.
S114, carrying out smooth denoising on the 2D laser radar data by adopting a mobile least square method;
S115, outputting the 2D laser radar data after noise reduction.
Example 2
A system for 2D lidar point cloud identification of pedestrians, comprising:
the data acquisition module is used for acquiring 2D laser radar data;
the data processing module is used for clustering the 2D laser radar data to obtain radar point cloud clusters;
the first judging module is used for screening out first candidate point cloud clusters according to the curvature of the radar point cloud clusters and the distance from the radar center;
the second judging module is used for screening out a second candidate point cloud cluster which meets a threshold value with the curvature of the first candidate point cloud cluster and the distance from the center of the radar and is closest to the first candidate point cloud cluster;
and the data output module is used for forming a pedestrian by the first candidate point cloud cluster and the second candidate point cloud cluster.
The method adopts the data acquisition module to acquire 2D laser radar data, and specifically, the data acquisition module is a 2D laser radar peripheral; clustering the 2D laser radar data by adopting a data processing module to obtain radar point cloud clusters, wherein the data processing module is an IMU processor; calculating and screening out first candidate point cloud clusters by adopting a first judging module, wherein the first judging module is an IMU processor; calculating and screening out second candidate point cloud clusters by adopting a second judging module, wherein the second judging module is an IMU processor; and the data output module combines and outputs the first candidate point cloud cluster and the second candidate point cloud cluster into a pedestrian, and outputs the pedestrian to the value controller for controlling the real-time path planning adjustment of the robot.
The application need not to adopt 3D laser radar to survey, can practice thrift a large amount of equipment input cost, and, 2D laser radar is because the data volume is few, be convenient for calculate and handle, and adopt the screening of degree of depth neural network to carry out the pedestrian in the prior art generally, then seriously increased the power burden, and 2D laser radar can accurately calculate the position of pedestrian through the abscissa of every laser radar point cloud, according to the position real-time control robot removal of pedestrian, accomplish the pedestrian and keep away the barrier, this application can accurately control the robot and keep away the barrier, facility input cost has been saved greatly under the limited circumstances of hardware resource. In some embodiments, functions or modules included in an apparatus provided by the embodiments of the present disclosure may be used to perform a method described in the foregoing method embodiments, and specific implementation of the method may refer to the description of the foregoing method embodiments, which is not repeated herein for brevity.
Example 3
An electronic device, comprising: the system comprises a chip, a processor and a memory, wherein the memory is used for storing computer program codes, the computer program codes comprise computer instructions, and the electronic equipment executes a method for identifying pedestrians by using 2D laser radar point clouds under the condition that the chip executes the computer instructions.
The electronic equipment comprises a processor, a memory, an input device and an output device. The processor, memory, input device, and output device are coupled by connectors, including various interfaces, transmission lines, buses, etc., as not limited in this application. It should be understood that in various embodiments of the present application, coupled is intended to mean interconnected by a particular means, including directly or indirectly through other devices, e.g., through various interfaces, transmission lines, buses, etc.
The processor may be one or more Graphics Processors (GPUs), which may be single-core GPUs or multi-core GPUs in the case where the processor is a GPU. Alternatively, the processor may be a processor group formed by a plurality of GPUs, and the plurality of processors are coupled to each other through one or more buses. In the alternative, the processor may be another type of processor, and the embodiment of the present application is not limited.
The memory may be used for storing computer program instructions, as well as various types of computer program code for executing the program code of the present application. Optionally, the memory includes, but is not limited to, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or portable read-only memory (CD-ROM) for the associated instructions and data.
The input means is for inputting data and/or signals and the output means is for outputting data and/or signals. The output device and the input device may be separate devices or may be a single device.
Example 4
A computer readable storage medium having stored therein a computer program comprising program instructions which, when executed by a processor of an electronic device, cause the processor to perform a method of 2D lidar point cloud recognition of pedestrians.
According to the invention, the 2D laser radar is adopted to externally acquire the point cloud for identifying pedestrians in the running process of the robot, the 2D laser radar can acquire the two-dimensional point cloud image, namely, the point cloud image is formed on the walking plane of the robot, and because the path planning is required to be carried out in real time based on the situation of pedestrians on the road surface in the walking process of the robot, the 2D laser radar can meet the requirement of pedestrian obstacle avoidance, the 3D laser radar is not required to be adopted for detection, a large amount of equipment throwing cost can be saved, and the 2D laser radar is convenient to calculate and process due to the small data quantity, in the prior art, the depth neural network is adopted to screen the pedestrians, so that the calculation force burden is seriously increased, the 2D laser radar can accurately calculate the positions of the pedestrians through the transverse and longitudinal coordinates of the point cloud of each laser radar, and the robot is controlled to move in real time according to the positions of the pedestrians, so that the pedestrian obstacle avoidance is completed.
The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for identifying pedestrians by using a 2D lidar point cloud, comprising:
acquiring 2D laser radar data;
clustering the 2D laser radar data to obtain radar point cloud clusters;
screening out a first candidate point cloud cluster according to the curvature of the radar point cloud cluster and the distance from the radar center;
screening out a second candidate point cloud cluster which has the curvature with the first candidate point cloud cluster and the distance from the radar center meeting a threshold and is closest to the first candidate point cloud cluster;
the first candidate point cloud cluster and the second candidate point cloud cluster form a pedestrian.
2. The method for identifying pedestrians by using 2D lidar point clouds according to claim 1, wherein the clustering the 2D lidar data to obtain radar point clouds comprises:
Sequentially comparing the distances of two continuous radar points according to the index sequence of the radar data, and when the distance between the two radar points exceeds a threshold value, considering that the two radar points do not belong to the same radar point cloud cluster;
after the radar point cloud is formed, judging whether the number of points in the radar point cloud meets the screening requirement, and when the number of points in the radar point cloud is smaller than or larger than a threshold value, eliminating the radar point cloud which does not meet the requirement.
3. The method for identifying pedestrians by using 2D lidar point clouds according to claim 1, wherein the step of screening out the first candidate point cloud according to the curvature of the radar point cloud and the distance from the radar center comprises:
calculating curvature of the radar point cloud: inputting coordinates of a starting point, a center point and an end point of the radar point cloud;
calculating the distance difference between the starting point abscissa and the intermediate point abscissa;
calculating the distance difference between the ordinate of the starting point and the ordinate of the intermediate point;
calculating the distance difference between the abscissa of the ending point and the abscissa of the intermediate point;
calculating the distance difference between the ordinate of the end point and the ordinate of the middle point;
calculating a first arctangent value of a distance difference between the starting point ordinate and the intermediate point ordinate divided by a distance difference between the starting point abscissa and the intermediate point abscissa;
Calculating a second arctangent value of the difference between the distances of the ending point ordinate and the intermediate point ordinate divided by the difference between the distances of the ending point abscissa and the intermediate point abscissa;
the curvature value is the difference between the first and second arctangent values plus 3.14, multiplied by 6.28%, and finally subtracted by the absolute value of 3.14.
4. A method of identifying pedestrians by a 2D lidar point cloud according to claim 3, wherein the screening out a first candidate point cloud based on the curvature of the radar point cloud and the distance from the radar center comprises:
calculating the distance between the radar point cloud and the radar center, and when the distance between the radar point cloud and the radar center is larger than a preset distance, conforming the current radar point cloud to the screening requirement of the first candidate point cloud;
and when one radar point cloud group meets the curvature screening requirement and the distance screening requirement simultaneously, marking the current radar point cloud group as a first candidate point cloud group.
5. The method of claim 1, wherein screening out a second candidate point cloud closest to the first candidate point cloud that meets a threshold with respect to a curvature of the first candidate point cloud and a distance from a radar center comprises:
After the first candidate point cloud cluster is selected, continuing to screen the radar point cloud clusters which are not screened;
judging the distance between the current radar point cloud and the first candidate point cloud when the difference between the curvature of the radar point cloud and the curvature of the first candidate point cloud and the difference between the distance between the radar point cloud and the radar center of the first candidate point cloud and the distance between the radar point cloud and the radar center of the first candidate point cloud are smaller than a threshold value;
and when the distance between the current radar point cloud and the first candidate point cloud is smaller than a threshold value, marking the current radar point cloud as a second candidate point cloud.
6. The method for identifying pedestrians by using 2D laser radar point clouds according to claim 2, wherein comparing distances between two consecutive radar points in order of the index of the radar data, and when the distance between two radar points exceeds a threshold value, considering that the two radar points do not belong to the same radar point cloud comprises:
acquiring coordinates of a radar point P1 and a radar point P2;
calculating a distance d between coordinates of the radar point P1 and the radar point P2;
when d is greater than 0.32, two radar points do not belong to the same radar point cloud, and when d is less than 0.32, two radar points belong to the same radar point cloud.
7. The method for identifying pedestrians by using 2D laser radar point cloud according to claim 1, wherein after the 2D laser radar data is acquired, the method further comprises denoising the 2D laser radar data, specifically:
obtaining parameters of Eps and MinPts in a DBSCAN algorithm;
establishing Kd-tree acceleration neighborhood search, and performing DBSCAN clustering denoising on 2D laser radar data noise;
setting the order of a moving least square method, carrying out normal estimation, and determining a fitting radius;
smoothing and denoising the 2D laser radar data by adopting a mobile least square method;
and outputting the noise-reduced 2D laser radar data.
8. A system for identifying pedestrians by a 2D lidar point cloud, comprising:
the data acquisition module is used for acquiring 2D laser radar data;
the data processing module is used for clustering the 2D laser radar data to obtain radar point cloud clusters;
the first judging module is used for screening out first candidate point cloud clusters according to the curvature of the radar point cloud clusters and the distance from the radar center;
the second judging module is used for screening out a second candidate point cloud cluster which meets a threshold value with the curvature of the first candidate point cloud cluster and the distance from the center of the radar and is closest to the first candidate point cloud cluster;
And the data output module is used for forming a pedestrian by the first candidate point cloud cluster and the second candidate point cloud cluster.
9. An electronic device, comprising: a chip, a processor and a memory for storing computer program code comprising computer instructions, which, when executed by the chip, the electronic device performs a method of identifying pedestrians by a 2D lidar point cloud according to any of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program comprising program instructions which, when executed by a processor of an electronic device, cause the processor to perform a method of identifying pedestrians by a 2D lidar point cloud according to any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410270594.1A CN117890882A (en) | 2024-03-11 | 2024-03-11 | Method and system for identifying pedestrians by 2D laser radar point cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410270594.1A CN117890882A (en) | 2024-03-11 | 2024-03-11 | Method and system for identifying pedestrians by 2D laser radar point cloud |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117890882A true CN117890882A (en) | 2024-04-16 |
Family
ID=90651999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410270594.1A Pending CN117890882A (en) | 2024-03-11 | 2024-03-11 | Method and system for identifying pedestrians by 2D laser radar point cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117890882A (en) |
-
2024
- 2024-03-11 CN CN202410270594.1A patent/CN117890882A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xia et al. | Geometric primitives in LiDAR point clouds: A review | |
| CN109509210B (en) | Obstacle tracking method and device | |
| CN112767490B (en) | Outdoor three-dimensional synchronous positioning and mapping method based on laser radar | |
| CN102248947B (en) | Object and vehicle detecting and tracking using a 3-D laser rangefinder | |
| CN111492265A (en) | Multi-resolution, simultaneous localization and mapping based on 3D lidar measurements | |
| JP2020514876A (en) | Apparatus, method and system for alignment of 3D data sets | |
| JP6456141B2 (en) | Generating map data | |
| CN108875804A (en) | A kind of data processing method and relevant apparatus based on laser point cloud data | |
| CN111612728A (en) | 3D point cloud densification method and device based on binocular RGB image | |
| CN112166457A (en) | Point cloud segmentation method and system and movable platform | |
| Choe et al. | Fast point cloud segmentation for an intelligent vehicle using sweeping 2D laser scanners | |
| CN111739099B (en) | Falling prevention method and device and electronic equipment | |
| CN114663526A (en) | Obstacle detection method, obstacle detection device, robot and computer-readable storage medium | |
| CN111736167B (en) | Method and device for obtaining laser point cloud density | |
| CN116379915A (en) | Building mapping method, device, system and storage medium | |
| CN111709988A (en) | Method and device for determining characteristic information of object, electronic equipment and storage medium | |
| CN116109601A (en) | Real-time target detection method based on three-dimensional laser radar point cloud | |
| CN113160292B (en) | Laser radar point cloud data three-dimensional modeling device and method based on intelligent mobile terminal | |
| CN113741523A (en) | Hybrid unmanned aerial vehicle autonomous detection method based on boundary and sampling | |
| CN113536959A (en) | Dynamic obstacle detection method based on stereoscopic vision | |
| CN117890882A (en) | Method and system for identifying pedestrians by 2D laser radar point cloud | |
| CN116468950A (en) | Three-dimensional target detection method for neighborhood search radius of class guide center point | |
| Spampinato et al. | Deep learning localization with 2D range scanner | |
| CN116863325A (en) | Method for multiple target detection and related product | |
| CN114384486A (en) | Data processing method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |