A laser radar-based charging pile identification method and mobile robot

Abstract

This invention provides a charging pile identification method and a mobile robot based on LiDAR. The method includes the following steps: S1, acquiring LiDAR data of the current position of the mobile robot and converting the LiDAR data into point cloud data; S2, downsampling the point cloud data to obtain a first region of interest (ROI) point cloud subset; S3, processing the first ROI point cloud subset to obtain at least one second ROI point cloud subset; S4, performing point cloud segmentation on each second ROI point cloud subset to obtain a first line subset, and filtering the first line subset to obtain at least one third ROI point cloud subset; S5, performing point cloud matching on each third ROI point cloud subset. This invention eliminates the need to change the algorithm even when the characteristics of the charging pile change, significantly shortening the identification time.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of robots, and particularly relates to a charging pile recognition method based on a laser radar and a mobile robot. BACKGROUND

[0002] With the continuous maturity of mobile robot technology, the autonomous return charging problem of the mobile robot is particularly important, and the charging pile recognition is the most important in the autonomous return charging. At present, a considerable part of the automatic return charging technology is completed by using an infrared sensor, and a small part of the automatic return charging technology is completed by using a laser radar. The status of similar technologies: for some methods of completing the autonomous return charging by using the laser radar, most of them are completed by searching for the features of the charging pile according to the laser data, and once the features of the charging pile change, the required algorithm needs to be changed. SUMMARY

[0003] In order to overcome the above technical defects, the application provides a charging pile recognition method based on a laser radar and a mobile robot, which can recognize different charging pile features without changing the algorithm.

[0004] The application is realized by the following scheme:

[0005] A charging pile recognition method based on a laser radar, comprising the following steps:

[0006] S1, acquiring laser data of a current position of a mobile robot, and converting the laser data into point cloud data;

[0007] S2, performing down-sampling processing on the point cloud data to obtain a first region of interest point cloud subset;

[0008] S3, processing the first region of interest point cloud subset to obtain at least one second region of interest point cloud subset;

[0009] S4, performing point cloud segmentation processing on each second region of interest point cloud subset to obtain a first straight line subset, and performing data screening on the first straight line subset to obtain at least one third region of interest point cloud subset;

[0010] S5, performing point cloud matching on each third region of interest point cloud subset.

[0011] As a further improvement of the application, the step S3 comprises:

[0012] S31, using point cloud clustering to extract point clouds from the first region of interest point cloud subset, filtering the first region of interest by setting the range of the number of clustered point clouds, and obtaining at least one second region of interest point cloud subset after clustering.

[0013] As a further improvement of the application, the step S31 comprises:

[0014] S311, selecting a seed point from the first point cloud subset of the region of interest;

[0015] S312, performing a radius neighborhood search on the seed point using a k-dimensional tree;

[0016] S313, if there is a point in the neighborhood, then the seed point is classified into the same cluster, and returning to step S311;

[0017] S314, if there is no point in the neighborhood, then the clustering ends;

[0018] If the cluster meets the threshold interval, the cluster is saved;

[0019] If the first point cloud subset of the region of interest is completed, at least one second point cloud subset of the region of interest is obtained;

[0020] If the first point cloud subset of the region of interest is not completed, returning to step S311.

[0021] As a further improvement of the application, the step S4 comprises:

[0022] S41, dividing each second point cloud subset of the region of interest into a plurality of straight lines, and saving the indexes of the plurality of straight lines as a first straight line subset.

[0023] As a further improvement of the application, the step S41 comprises:

[0024] S411, selecting a point cloud cluster of the second point cloud subset of the region of interest;

[0025] S412, performing straight line extraction on the point cloud cluster to obtain a first straight line subset;

[0026] S413, all straight lines in the first straight line subset meeting a threshold condition;

[0027] S414, if the second point cloud subset of the region of interest is completed, at least one third point cloud subset of the region of interest is obtained;

[0028] S415, if the second point cloud subset of the region of interest is not completed, returning to step S411.

[0029] As a further improvement of the application, the step S413 comprises:

[0030] S4131, if the number of straight lines in the first straight line subset is less than 2, selecting a straight line in the first straight line subset;

[0031] S4132, if the angle difference between the straight line and other straight lines in the first straight line subset meets the threshold condition, performing threshold condition judgment on another straight line until the traversal of the straight lines in the first straight line subset is completed; if the angle difference between the straight line and other straight lines in the first straight line subset does not meet the threshold condition, deleting the point cloud cluster from the second point cloud subset in the region of interest.

[0032] As a further improvement of the application, the step S413 comprises:

[0033] S4133, if the number of straight lines in the first straight line subset is greater than or equal to 2, deleting the point cloud cluster from the second point cloud subset in the region of interest.

[0034] 8. The charging pile recognition method, wherein the step S5 comprises:

[0035] S51, obtaining a point cloud cluster from the third point cloud subset in the region of interest;

[0036] S52, performing icp matching between the point cloud cluster and the charging template;

[0037] S53, updating the lowest icp score;

[0038] S54, if the traversal of the third point cloud subset in the region of interest is completed, obtaining a transformation matrix according to the lowest icp score and calculating the pose of the charging pile;

[0039] S55, if the traversal of the third point cloud subset in the region of interest is not completed, returning to step S51.

[0040] In addition, the application further provides a mobile robot, characterized by comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to realize the charging pile recognition method.

[0041] Compared with the prior art, the present application has the following beneficial effects: when the mobile robot navigates to the vicinity of the charging pile, the mobile robot starts to identify the charging pile, first the mobile robot acquires the data of the laser radar at the current position of the mobile robot, and converts the data of the laser radar into point cloud data. Since the point cloud data may have some noise, the point cloud data is first subjected to a downsampling operation, and the point cloud data after the downsampling operation is defined as a first region of interest point cloud subset; then the first region of interest point cloud subset is subjected to point cloud clustering segmentation processing, the range of the clustered point cloud is set according to the size of the charging pile, the characteristics of the laser radar and the like, and a plurality of second region of interest point cloud subsets (including the point cloud of the charging pile part) are effectively extracted; then the extracted second region of interest point cloud subset is subjected to data screening according to the characteristics of the charging pile, and the point cloud more conforming to the characteristics of the charging pile is further screened out, and at least one third region of interest point cloud subset is obtained; finally, the point cloud matching is performed on each third region of interest point cloud subset, and the accurate position of the charging pile is matched. By this method, when the characteristics of the charging pile change, the algorithm also does not need to be changed, and the identification time is greatly shortened. BRIEF DESCRIPTION OF DRAWINGS

[0042] The specific embodiments of the present application will be further described in detail below with reference to the accompanying drawings, in which:

[0043] Figure 1 Flowchart of the charging pile identification method described in Embodiment 1;

[0044] Figure 2 Flowchart of step S3 in the charging pile identification method described in Embodiment 1;

[0045] Figure 3 Flowchart of step S4 in the charging pile identification method described in Embodiment 1;

[0046] Figure 4 Flowchart of step S5 in the charging pile identification method described in Embodiment 1.

[0047] Figure 5 Charging pile template described in Embodiment 1. DETAILED DESCRIPTION

[0048] The preferred embodiments of the present application will be described below with reference to the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0049] It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, the sequence numbers of each step are merely used to distinguish between steps and do not imply that each step must be strictly performed in the order of the numbers.

[0050] Definitions:

[0051] Point cloud data refers to a collection of vectors in a three-dimensional coordinate system. Besides geometric location, point cloud data sometimes also contains color information. Color information is typically obtained by acquiring color images from a camera and then assigning the color information (RGB) of the corresponding pixels to the corresponding points in the point cloud. Intensity information is obtained from the echo intensity collected by a laser scanner receiver. This intensity information is related to the target's surface material, roughness, incident angle, and the instrument's emission energy and laser wavelength.

[0052] Point cloud clustering algorithm: used to extract point clouds.

[0053] Example 1

[0054] This invention provides a charging pile identification method based on lidar, such as... Figure 1 As shown, the steps include:

[0055] S1. When the robot receives a return-to-charge command or needs to return to charge autonomously due to insufficient power, the robot will navigate to the vicinity of the charging station based on the pre-set charging station location. After navigation is completed, the robot will acquire the laser data of its current location and convert the laser data into point cloud data.

[0056] S2. Since the data collected by the lidar contains noise, which can cause irregular point cloud data density, it is necessary to downsample the point cloud data. The point cloud data obtained by downsampling is defined as a subset of the point cloud of the first region of interest.

[0057] S3. Process the point cloud subset of the first region of interest to obtain at least one point cloud subset of the second region of interest.

[0058] Specifically, such as Figure 2 As shown, step S3 includes:

[0059] S31. Point cloud clustering is used to extract point clouds from the first region of interest (ROI) subset. The ROI is filtered by setting a range for the number of clustered point clouds, resulting in at least one second ROI subset. Euclidean clustering is then performed on the first ROI subset. After Euclidean clustering, several second ROI subsets (including the point clouds from the charging pile section) can be effectively extracted.

[0060] Further, step S31 includes:

[0061] S311. Select seed points from the first subset of the point cloud of the region of interest.

[0062] S312. Use a k-dimensional tree to perform a neighborhood search of the seed point with radius r.

[0063] S313. If there are points in the neighborhood, the seed points are assigned to the same cluster Q, and the process returns to step S311 to continue selecting seed points from the first region of interest point cloud subset.

[0064] S314. If there are no points in the neighborhood, that is, if one of the seed points cannot be found within the radius of r, then the clustering of cluster Q ends.

[0065] The closer the robot is to the charging station, the more points the charging station will have, and vice versa. The smaller the angular resolution of the LiDAR, the more points the LiDAR will generate in one rotation, resulting in more points corresponding to the charging station. The larger the charging station, the more points the LiDAR will hit. Therefore, the threshold range [Ec_min_size, Ec_max_size] for the number of points in the clustered point cloud is determined based on the robot's approximate distance to the charging station, the LiDAR's angular resolution, and the size of the charging station. If the number of points in cluster Q is within the threshold range, the clustering result is saved.

[0066] If the first region of interest point cloud subset is traversed, then at least one second region of interest point cloud subset is obtained.

[0067] If the first region of interest (ROI) point cloud subset has not been traversed, return to step S311 and continue selecting seed points from the first ROI point cloud subset until the first ROI point cloud subset has been traversed.

[0068] S4. Perform point cloud segmentation on each second region of interest point cloud subset to obtain a first straight line subset. Filter the first straight line subset to obtain at least one third region of interest point cloud subset.

[0069] Specifically, as shown in Figure 3, step S4 includes:

[0070] S41. Divide each subset of the point cloud of the second region of interest into several straight lines, and save the indices of the several straight lines as the first subset of straight lines.

[0071] Further, step S41 includes:

[0072] S411. Select a point cloud cluster from the second region of interest.

[0073] S412. Perform RANSAC line extraction on the point cloud clusters. Each point cloud cluster can obtain at least one line, and these lines constitute the first subset of lines.

[0074] S413. In the first subset of straight lines, all straight lines that satisfy the threshold condition.

[0075] Further, step S413 includes:

[0076] S4131. If the number of lines in the first subset of lines is less than 2, select one line from the first subset of lines.

[0077] S4132. If the angle difference between the line and other lines in the first line subset satisfies the threshold condition [angle_min, angle_max], then the threshold condition is judged for another line until the traversal of the lines in the first line subset is completed; if the angle difference between the line and other lines in the first line subset does not satisfy the threshold condition, then the point cloud cluster is deleted from the second region of interest point cloud subset.

[0078] S4133. If the number of lines in the first straight line subset is greater than or equal to 2, then the point cloud cluster is deleted from the second point cloud subset of interest.

[0079] S414. If the second region of interest point cloud subset has been traversed, then at least one third region of interest point cloud subset is obtained.

[0080] S415. If the traversal of the second region of interest point cloud subset is not completed, return the value from step S411.

[0081] S5. Perform point cloud matching for each third region of interest subset of the point cloud.

[0082] Specifically, as shown in Figure 4, step S5 includes:

[0083] S51. Obtain any point cloud cluster from the third region of interest point cloud subset.

[0084] S52. Cluster the point cloud and match it with the charging template, which is prepared in advance according to the shape of the charging pile.

[0085] S53. Update the lowest score of ICP, where ICP is the Iterative Nearest Point Cloud Matching Algorithm.

[0086] S54. If the third region of interest point cloud subset has been traversed, the transformation matrix is ​​obtained based on the lowest ICP score, and the pose of the charging pile is calculated.

[0087] S55. If the traversal of the third region of interest point cloud subset is not completed, return to step S51.

[0088] Therefore, this embodiment has the following beneficial effects:

[0089] 1. Point cloud clustering segmentation can effectively segment the clusters of point clouds of interest. The threshold range of the number of points in the clustered point cloud is mainly determined based on factors such as the approximate position of the robot from the charging station, the angular resolution of the LiDAR, and the size of the charging station.

[0090] 2. Extracting straight lines from clustered point clouds can better filter out other clustered point cloud clusters that do not belong to charging piles, effectively improving the accuracy of identification and speeding up the identification time.

[0091] 3. No algorithm changes are needed; simply changing the charging pile template to different shapes is sufficient to identify the features of various charging piles.

[0092] Example 2

[0093] This embodiment provides a mobile robot, including a processor and a memory. The memory stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement the charging pile identification method in Embodiment 1. The mobile robot can be a sweeping robot.

[0094] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0095] The memory can be used to store the computer program or module. The processor implements various functions of the charging pile identification method by running or executing the computer program or module stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application required for a function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0096] The above are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for identifying a charging pile based on a laser radar, characterized in that, The method comprises the steps of: S1, acquiring laser data of the current position of the mobile robot, and converting the laser data into point cloud data; S2, performing down-sampling processing on the point cloud data to obtain a first region-of-interest point cloud subset; S3, processing the first region-of-interest point cloud subset to obtain at least one second region-of-interest point cloud subset: using point cloud clustering to extract point clouds from the first region-of-interest point cloud subset, filtering the first region-of-interest by setting a range of the number of clustered point clouds, and obtaining at least one second region-of-interest point cloud subset after clustering; S4, performing point cloud segmentation processing on each second region-of-interest point cloud subset to obtain a first straight line subset, and performing data screening on the first straight line subset to obtain at least one third region-of-interest point cloud subset: S41, dividing each second region-of-interest point cloud subset into a plurality of straight lines, and saving the indexes of the plurality of straight lines as the first straight line subset: S411, selecting a point cloud cluster of the second region-of-interest point cloud subset; S412, extracting a straight line from the point cloud cluster to obtain the first straight line subset; S413, in the first straight line subset, all straight lines satisfying a threshold condition: S4131, if the number of straight lines in the first straight line subset is less than 2, selecting a straight line in the first straight line subset; S4132, if the angle difference between the straight line and other straight lines in the first straight line subset satisfies a threshold condition, performing threshold condition judgment on another straight line until the traversal of the straight lines in the first straight line subset is completed; if the angle difference between the straight line and other straight lines in the first straight line subset does not satisfy the threshold condition, deleting the point cloud cluster from the second region-of-interest point cloud subset S414, if the traversal of the second region-of-interest point cloud subset is completed, obtaining at least one third region-of-interest point cloud subset; S415, if the traversal of the second region-of-interest point cloud subset is not completed, returning to step S411; S5, performing point cloud matching on each third region-of-interest point cloud subset. 2.The charging pile recognition method of claim 1, characterized in that, The step S3 comprises: S311, selecting a seed point from the first region-of-interest point cloud subset; S312, performing radius neighborhood search on the seed point by using a k-dimensional tree; S313, if there is a point in the neighborhood, the seed point is classified into the same cluster, and the process returns to step S311; S314, if there is no point in the neighborhood, the clustering is completed; if the cluster satisfies a threshold interval, the cluster is saved; if the traversal of the first region-of-interest point cloud subset is completed, at least one second region-of-interest point cloud subset is obtained; if the traversal of the first region-of-interest point cloud subset is not completed, the process returns to step S311. 3.The charging pile recognition method of claim 1, characterized in that, The step S413 comprises: S4133, if the number of straight lines in the first straight line subset is greater than or equal to 2, the point cloud cluster is deleted from the second region-of-interest point cloud subset. 4.The charging pile recognition method of claim 3, characterized in that, The step S5 comprises: S51, acquiring a point cloud cluster from the third region-of-interest point cloud subset; S52, performing icp matching between the point cloud cluster and a charging template; S53, updating the lowest icp score; S54, if the traversal of the third region-of-interest point cloud subset is completed, obtaining a transformation matrix according to the lowest icp score, and calculating the pose of the charging pile; S55, if the third point cloud subset of the region of interest is not traversed, return to step S51.

5. A mobile robot, characterized by The device comprises a processor and a memory, and the memory stores at least one instruction, at least one program, a code set or an instruction set, which is loaded and executed by the processor to realize the charging pile identification method in any one of claims 1-4.

Images