A method for filtering out sunlight noise in a surface mine area by fusing multiple laser radars
By setting up multiple lidar units at different locations and angles in the open-pit mine, and utilizing cross-validation and cluster analysis, the noise problem caused by direct sunlight was solved, enabling accurate detection of obstacles and effective filtering of sunlight noise.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TAGE IDRIVER TECHNOLOGY CO LTD
- Filing Date
- 2023-03-27
- Publication Date
- 2026-06-09
AI Technical Summary
LiDAR in open-pit mines is prone to generating large areas of sunlight noise when exposed to direct sunlight, which affects autonomous driving's judgment of obstacles. Existing technologies are unable to effectively filter out these noises.
By combining multiple lidars installed at different locations and angles, and through cross-validation and cluster analysis, obstacle point clouds and sunlight noise point clouds are distinguished. Sunlight noise is removed by using multi-liquidity cross-validation, while ensuring the accuracy of obstacle detection.
It effectively removes sunlight noise, improves the accuracy and robustness of lidar obstacle detection in open-pit mines, and reduces the false detection rate.
Smart Images
Figure CN116299557B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of lidar point cloud noise reduction technology, and more specifically, to a method for filtering sunlight noise points in open-pit mines using multi-lidar fusion. Background Technology
[0002] LiDAR (Light Detection and Ranging) measures the propagation distance between the sensor transmitter and the target object, analyzing information such as the magnitude, amplitude, frequency, and phase of the reflected energy, and the reflected spectrum, to present precise three-dimensional structural information of the target object in the form of a point cloud. It is a crucial sensor in the field of autonomous driving.
[0003] With the development of autonomous driving, LiDAR technology is also rapidly iterating. Existing LiDAR mainly includes solid-state LiDAR and mechanical LiDAR. Solid-state LiDAR has high resolution but a limited horizontal field of view; mechanical LiDAR has a 360-degree horizontal field of view, but its resolution decreases significantly with increasing distance.
[0004] Extensive comparative testing of products on the market revealed that direct sunlight causes large areas of sunlight noise near the LiDAR sensor, while this issue does not occur under indirect sunlight. Due to the rugged terrain, varied slopes, and lack of building obstructions in open-pit mines, autonomous vehicles experience direct sunlight for more than two hours daily, resulting in significant sunlight noise. This extensive sunlight noise negatively impacts the autonomous driving system's ability to determine the presence or absence of obstacles.
[0005] Existing technologies filter sunlight noise at the radar echo signal level; however, commercially available products still produce sunlight noise that cannot be filtered out. Therefore, post-processing at the laser point cloud level is necessary. Existing post-processing methods filter sunlight noise by sampling sunlight noise and performing region filtering, or by directly performing discrete point filtering. These methods do not consider situations where the noise contains obstacles or large areas of continuous noise.
[0006] Therefore, a method for filtering sunlight noise in open-pit mines using multi-lidar fusion is proposed to solve the aforementioned problems. Summary of the Invention
[0007] The present invention aims to provide a method for filtering sunlight noise in open-pit mines by fusing multiple lidar sensors, in order to solve or improve the problem that lidar sensors easily generate large areas of sunlight noise in a fixed area when exposed to direct sunlight.
[0008] In view of this, the first aspect of the present invention is to provide a method for filtering sunlight noise in open-pit mines using multi-lidar fusion.
[0009] The first aspect of the present invention provides a method for filtering sunlight noise in open-pit mines using multi-LiDAR fusion. This method involves setting up a first LiDAR and a second LiDAR at different installation positions and / or angles. The scanning area of the second LiDAR can cover the region R in which dense sunlight noise is generated by the first LiDAR. The method is characterized by its ability to distinguish between obstacle point clouds and sunlight noise point clouds. The method includes the following steps: Step 1, acquiring point clouds from the first and second LiDARs within the region R and merging them into a dataset; Step 2, clustering the dataset to obtain N point cloud clusters, and determining whether each point cloud cluster containing the point cloud from the first LiDAR contains the second LiDAR. Point cloud of lidar; Scenario 1: When the point cloud cluster with the first lidar contains the point cloud of the second lidar, compare the density of the point clouds of the first lidar and the second lidar within the point cloud cluster and perform adaptation, then calculate the similarity between the point clouds of the first lidar and the second lidar; Scenario 2: When the point cloud cluster with the first lidar does not contain the point cloud of the second lidar, calculate the parameters of the point cloud cluster and the parameters of the two beams at the front and rear ends of the emission range of the second lidar located at the center of the point cloud cluster, as well as the proportion of effective point cloud inside the point cloud cluster; Step 3: Based on the results calculated in Step 2, classify the point cloud of the first lidar into obstacle point cloud or sunlight noise point cloud.
[0010] This invention provides a method for filtering sunlight noise in open-pit mines using multi-LiDAR fusion. Based on the characteristic that LiDAR easily generates large-area sunlight noise in a fixed area when sunlight is direct, but does not generate this phenomenon when there is no direct sunlight, the method uses multiple LiDAR verification to determine the detection consistency of LiDAR point clouds in public areas, thereby ensuring the detection of obstacles and filtering out false detections of sunlight noise.
[0011] It can not only remove the sunlight noise detected by scanning, but also accurately identify the obstacles detected by scanning. It can solve the problem of obstacles being included in the sunlight noise in complex environmental scanning results, and it is more robust in actual vehicle operation.
[0012] In addition, the technical solutions provided by embodiments of the present invention may also have the following additional technical features:
[0013] In any of the above technical solutions, before step two, step one further includes: deleting the ground point cloud from the dataset; determining whether the dataset is an empty set; and retaining the dataset as a non-empty set.
[0014] In this technical solution, in order to ensure the smooth implementation of subsequent steps and data simplification, the ground point cloud in the synthesized dataset is deleted to reduce the internal data volume and improve the speed of subsequent calculation processes. At the same time, for datasets containing only ground point clouds, the data is deleted after the ground point clouds are cleared, thereby further reducing the data sample and retaining datasets that only include non-ground point clouds.
[0015] In any of the above technical solutions, in scenario one of step two: when the point cloud density of the first lidar is greater than that of the second lidar, a circumferential beamline is constructed by adding or subtracting offsets at the beamline positions of the second lidar's point cloud within the point cloud cluster, and the point cloud of the first lidar located within the beamline is extracted to form a new point cloud; or when the point cloud density of the first lidar is less than that of the second lidar, a circumferential beamline is constructed by adding or subtracting offsets at the beamline positions of the first lidar's point cloud within the point cloud cluster, and the point cloud of the second lidar located within the beamline is extracted to form a new point cloud; or when the point cloud densities of the first lidar and the second lidar are equal, the point cloud of either the first lidar or the second lidar is selected as the new point cloud; the point cloud distance between the new point cloud and the corresponding point cloud in the first lidar or the second lidar is calculated and used as the similarity.
[0016] In this technical solution, a vertical offset deta_z is added based on the position of the low-beam lidar beam to construct a beam band. For dense point clouds, the lidar point cloud inside the beam band is extracted and then similarity calculation is performed, thus eliminating the problems of low vertical resolution and significant variation with distance of the low-beam lidar.
[0017] With the second lidar removing sunlight noise from the first lidar, different approaches are adopted for different scenarios: the point cloud density of the first lidar is greater than that of the second lidar, the point cloud density of the first lidar is less than that of the second lidar, and the point cloud density of the first lidar and the second lidar are equal. One of these approaches selects a point cloud near the line band and uses it as a new point cloud for calculation with the other lidar. This avoids the problem of uneven point cloud density obtained by lidars at different distances, which leads to large deviations in the similarity calculation of corresponding points, and ensures the accuracy of subsequent calculations.
[0018] In any of the above technical solutions, step three includes: S301, determining the similarity between the similarity and the point cloud similarity threshold d. CD_threadshold Size; if similarity is less than d CD_threadshold If the similarity is greater than d, then the new point cloud is the point cloud of the obstacle. CD_threadshold The new point cloud then contains sunlight noise.
[0019] In this technical solution, the similarity is calculated by using the point cloud distance between the new point cloud and its corresponding point cloud, and then compared with d.CD_threadshold By comparing sizes, new point cloud types can be distinguished. It can also determine whether a point cloud in the same region R containing both a first lidar and a second lidar is a solar noise point cloud.
[0020] In any of the above technical solutions, in the second scenario of step two, the parameters of the point cloud cluster and the parameters of the two line beams at the front and rear ends of the emission range of the second lidar located at the center of the point cloud cluster are calculated as follows: Construct the bounding box of the point cloud cluster and obtain the parameters of the bounding box including length, width, height, and horizontal distance between the center point and the second lidar; calculate the horizontal distances S1 and S2 between the two line beams and the second lidar; calculate the ground distance between the two line beams and the maximum height of the obstacle that can be accommodated between the two line beams.
[0021] In any of the above technical solutions, step three further includes: S302, determining whether the length and width of the bounding frame are less than the ground spacing; if so, otherwise the point cloud of the point cloud cluster is a sunlight noise point cloud; if so, the point cloud of the point cloud cluster is an obstacle point cloud; or determining whether the sum of the horizontal distance between the center point and the second lidar and the length of the bounding frame is between S1 and S2; if so, otherwise the point cloud of the point cloud cluster is a sunlight noise point cloud; if so, the point cloud of the point cloud cluster is an obstacle point cloud; or determining whether the height of the bounding frame is less than the maximum height of the obstacle that can be accommodated between the two beams; if so, otherwise the point cloud of the point cloud cluster is a sunlight noise point cloud; if so, the point cloud of the point cloud cluster is an obstacle point cloud.
[0022] In this technical solution, the above calculation conditions are used to determine the first-level distinction between sunlight noise point clouds and obstacle point clouds based on the bounding box's length, width, height, and center point.
[0023] In any of the above technical solutions, the step of determining the proportion of effective point clouds within the point cloud cluster includes: taking each point of the obstacle point cloud as the origin and defining a search area with a radius of r; counting the search areas where the number of internal point clouds exceeds the point count threshold num_threadshold, and taking the origin corresponding to the search area as an effective point; and calculating the proportion of effective points in each search area to all point clouds in the point cloud cluster.
[0024] In this technical solution, the number of point clouds in a search area with a radius of r is statistically analyzed and compared to retain the search areas that meet the conditions. The point clouds of the origin inside the search area are then re-evaluated, resulting in a more accurate final evaluation result.
[0025] In any of the above technical solutions, step three further includes: S303, obtaining the effective point proportion threshold threadshold and comparing it with the proportion of the effective point cloud; if the proportion of the effective point cloud is large, the origin is a sunlight noise point cloud; if the proportion of the effective point cloud is small, the origin is an obstacle point cloud.
[0026] In this technical solution, a set threshold of the percentage of valid points (threadshold) is used to determine the proportion of points, and the point cloud at the corresponding origin is classified as either a sunlight noise point cloud or an obstacle point cloud based on the magnitude of the result.
[0027] The beneficial effects of this invention compared to the prior art are as follows:
[0028] By using a fusion method that cross-verifies multiple LiDARs, LiDARs installed at different angles can cross-verify each other, removing sunlight noise from each other's point clouds, while ensuring obstacle detection and reducing false deletions.
[0029] A method is proposed to construct a beamband based on the beam position of low-beam lidar by adding vertical offsets. For dense point clouds, lidar point clouds inside the beamband are extracted and similarity calculation is performed, which eliminates the problems of low vertical resolution and significant variation with distance of low-beam lidar.
[0030] By calculating the distance from the obstacle to the radar, the beams of two adjacent lidars near the obstacle are deduced. Multi-radar cross-verification is performed for cases where the obstacle size is larger than the beam gap and cases where the obstacle size is smaller than the beam gap. This method is compatible with the noise judgment of point cloud clusters that exceed the gap and point cloud clusters that are within the gap.
[0031] A method is proposed to use discrete point filtering to determine the number of point cloud clusters within the gap of a low-beam lidar system. At the same time, the composition ratio analysis is performed using the number of points that meet the conditions to qualitatively identify noise points.
[0032] Additional aspects and advantages of embodiments of the invention will become apparent in the following description or may be learned by practice of embodiments of the invention. Attached Figure Description
[0033] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.
[0034] Figure 1 This is a schematic diagram of the cross-validation of multiple lidar sensors in this invention;
[0035] Figure 2 This is a schematic diagram of the discrete point filtering of the present invention. Detailed Implementation
[0036] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.
[0037] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.
[0038] Please see Figure 1-2 The first aspect of the present invention provides a method for filtering sunlight noise in open-pit mines using multi-lidar fusion, comprising:
[0039] First, determine the area R where the first lidar generates dense sunlight noise, and the distance between area R and the lidar, specifically 5m-10m; then, install a second lidar, ensuring its acquisition range covers area R. The second lidar should be installed at an angle to the first lidar. Figure 1 (a) Assume that the noise of the first lidar is eliminated by using the second lidar.
[0040] Simultaneously acquire data from the first and second lidar sensors, and extract point cloud data Data1 from the first lidar sensor and Data2 from the second lidar sensor within region R; remove the ground point clouds from Data1 and Data2, and combine the remaining point cloud data of Data1 and Data2 into a set obj_data; determine whether obj_data is an empty set. If it is empty, it means that no obstacle was detected. If it is not empty, cluster the data in obj_data to generate N point cloud clusters cluster1, cluster2, ..., clusterN.
[0041] For each point cloud cluster N, cross-validation is used to determine the component affiliation and further to determine whether it is a point of sunlight noise.
[0042] (1) If points generated by both the first lidar and the second lidar exist simultaneously in a point cloud cluster, the point pairs generated by the first lidar and the second lidar are separated:
[0043] The point cloud cluster is divided into obj_cluster1 and obj_cluster2 according to the first lidar and the second lidar, and the similarity between obj_cluster1 and obj_cluster2 is calculated.
[0044] Case 1: If the first lidar is a dense point cloud lidar, such as a pixel-type solid-state lidar; and the second lidar is a low-beam lidar, such as a mechanical lidar; the point cloud in obj_cluster1 needs to be extracted according to the rings of obj_cluster2 corresponding to the second lidar with a height of 0.01 meters above and below the rings, to form obj_cluster_new1.
[0045] Case 2: If the first lidar is a low-beam lidar, such as a mechanical lidar; and the second lidar is a dense point cloud lidar, such as a pixel-type solid-state lidar; the point cloud in obj_cluster2 needs to be extracted according to the rings of obj_cluster1 corresponding to the first lidar, with a height of 0.01 meters above and below each ring, to form obj_cluster_new2.
[0046] Case 3: If the first lidar and the second lidar have the same beamline, calculate the chamfer distance between obj_cluster_new1 and obj_cluster2;
[0047] The matching points of the lidar are extracted based on the beamline, so as to eliminate the calculation error introduced by the rapid increase of the vertical spacing of the beamlines in low beamline lidar with increasing distance.
[0048] For example, in case 1, the following formula is used:
[0049]
[0050] Where, d CD (obj_cluster_new1,obj_cluster2) represents the point cloud distance between point clouds obj_cluster_new1 and obj_cluster2, and N is the distance between them. obj_cluster_new1 N represents the number of points in the point cloud obj_cluster_new1. obj_cluster2 Let p1 be the number of points in the point cloud obj_cluster2, p2 be any point in the point cloud obj_cluster_new1, and p3 be any point in obj_cluster2. It is the square of the L2 norm.
[0051] When d CD (obj_cluster_new1,obj_cluster2) <d CD_threadshold When the similarity meets the requirements, it is a point cloud on the same obstacle.
[0052] When d CD (obj_cluster_new1,obj_cluster2)>d CD_threadsholdWhen sunlight noise exists, calculate the bounding box length L_cluster2, width W_cluster2, and height H_cluster2 of the point cloud cluster obj_cluster2, and generate rectangular outline points based on the center point and length and width.
[0053] Extract the point cloud of obj_cluster1 within the bounding box of obj_cluster2: Determine whether the point cloud within obj_cluster1 falls within the rectangular outline formed by obj_cluster2 and its height is less than H_cluster2. If the condition is met, it is treated as obstacle point cloud, and the remaining obj_cluster1 points are sunlight noise.
[0054] (2) If there are points generated by the first lidar in the point cloud cluster, but no points generated by the second lidar.
[0055] Construct the bounding box of the point cloud cluster, and obtain the center point of the bounding box and the length L_cluster, width W_cluster and height H_cluster;
[0056] The horizontal distance L from the center point of the point cloud cluster to the second lidar and the included vertical angle theta are calculated using the following formula:
[0057]
[0058] Among them, the angle between the line connecting the center point of the theta point cloud cluster and the center of the second lidar and the vertical direction, and tan -1 (*) represents the arctangent function, L represents the horizontal distance L from the center point of the cloud cluster to the second lidar, and H represents the installation height of the center of the second lidar.
[0059] Based on the installation height H of the second lidar, and according to the installation tilt angle and the vertical field of view of the second lidar, the angle between the lowest beam of the second lidar and the vertical direction is determined using the following formula:
[0060]
[0061] Where β is the angle between the lowest beam of the second lidar and the vertical direction, α is the elevation angle of the second lidar (downward is negative, upward is positive), and FOV is... ⊥ This is the vertical field of view of the second lidar.
[0062] Based on the vertical resolution θ of the second lidar beam, the positions of the point cloud cluster center in the two lidar beams n1 and n2 before and after the second lidar are determined using the following formula:
[0063] n1 = Floor((theta-β) / θ)
[0064] n2 = ceil((theta - β) / θ)
[0065] Where ceil(*) is rounding up, floor(*) is rounding down, n1 and n2 are the beam numbers counted from the lowest beam of the second lidar and n1 < n2, n1 is the beam before the center of the point cloud cluster, and n2 is the beam after the center of the point cloud cluster. Thus, the horizontal distances S1 and S2 (S1 < S2) from the two beams to the second lidar are obtained, and the following formula is specifically used:
[0066] S1 = H * tan(β + n1 * θ)
[0067] S2 = H * tan(β + n2 * θ)
[0068] Where S1 is the horizontal distance from the beam n1 of the second lidar to the second lidar, S2 is the horizontal distance from the beam n2 of the second lidar to the second lidar, and tan(*) is the tangent function.
[0069] Calculate the ground spacing deta_r between the two beams, and at the same time determine the maximum height deta_h that allows the two beams to hit the horizontal ground. The following formula is specifically used:
[0070] deta_r = S2 - S1
[0071]
[0072] Judge whether the following conditions are met:
[0073] L_cluster < deta_r
[0074] && W_cluster1 < deta_r && (L + L_cluster) < S2 && (L + L_cluster) > S1 && H_cluster1 < deta_h;
[0075] If not, it is a sunlight noise point, such as Figure 1 (b) case does not satisfy L_cluster < deta_r, such as Figure 1 (d) case does not satisfy H_cluster1 < deta_h. If there are points generated by the first lidar and no points generated by the second lidar in this point cloud cluster cluster when the radar is working properly, it is qualitatively a sunlight noise point;
[0076] If satisfied, further use the radius outlier filter to judge its discreteness, such as Figure 2Traverse whether the number of adjacent points within the radius region r of each point is greater than the threshold num_threadshold. If satisfied, it is recorded as a valid point. The number of valid points N is divided by the total number of points in the point cloud cluster N_cluster. If it is less than the threshold, N / N_cluster < threadshold, it is qualitatively determined as noise; if it is greater than the threshold N / N_cluster > threadshold, it is retained as an obstacle.
[0077] Among them, the threshold needs to be set according to different types and brands of lidar.
[0078] Furthermore, for Data1, first filter the values near the maximum reflection intensity according to the reflection intensity of the point cloud. For example, the set range of the reflection intensity of the lidar is 0 - 255, and when the lidar hits an object, the reflection intensity of the generated points reaching 250 - 255 is very few. However, due to direct sunlight, there are many cases where the reflection intensity of the sunlight noise reaches 250 - 255. After filtering based on the reflection intensity, the proportion of the existence of sunlight noise will be reduced.
[0079] Through the method described in this embodiment, on the basis of ensuring obstacle detection, the interference of sunlight noise on the lidar is eliminated.
[0080] Embodiment 1
[0081] For scenarios where there are overlapping regions for three or more lidars (S1, S2, S3), for the point cloud clusters in the overlapping regions, calculate S1 and S2 according to the above method to obtain the determination of whether it is sunlight noise, noise_flag1; calculate S1 and S3 according to the above method to obtain the determination of whether it is sunlight noise, noise_flag2; calculate S2 and S3 according to the above method to obtain the determination of whether it is sunlight noise, noise_flag3. The optional values of noise_flag1, noise_flag2, and noise_flag3 are 0 or 1. Select 0 for the point cloud cluster as an obstacle, and select 1 for the point cloud cluster as sunlight noise. According to the characteristics of the lidar, set the determination ability weight w1 of S1 and S2, set the determination ability weight w2 of S1 and S3, and set the determination ability weight w3 of S2 and S3. If w1×noise_flag1 + w2×noise_flag2 + w3×noise_flag3 > 0.5, then the current point cloud cluster is sunlight noise.
[0082] Embodiment 2
[0083] For three or more lidars (S1, S2, S3), in the scenario where the combination of S2 and S3 has an overlapping region with S1, for the point cloud clusters in the overlapping regions, first merge the S2 point cloud cluster and the S3 point cloud cluster into a new point cloud cluster, denoted as M_23, and then calculate S1 and M_23 according to the above method to obtain the determination of whether it is sunlight noise;
[0084] Example 3
[0085] For scenarios involving three or more lidar units (S1, S2, S3, S4), where the combination of S2 and S3 overlaps with areas of S1 and S4, the point cloud clusters of S2 and S3 are preferentially merged into a new point cloud cluster, denoted as M_23. For point cloud clusters in the overlapping area, S1 and M_23 are calculated using the method described above to obtain the noise_flag1 indicating whether it is sunlight noise; S1 and S4 are calculated using the method described above to obtain the noise_flag2 indicating whether it is sunlight noise; and M_23 and S4 are calculated using the method described above to obtain the noise_flag3 indicating whether it is sunlight noise. The values for noise_flag1, noise_flag2, and noise_flag3 can be 0 or 1; 0 is selected for point cloud clusters representing obstacles, and 1 is selected for point cloud clusters representing sunlight noise. The determination capability weights w1 for S1 and M_23, w2 for S1 and S4, and w3 for M_23 and S4 can be set according to the lidar characteristics. If w1×noise_flag1+w2×noise_flag2+w3×noise_flag3>0.5, then the current point cloud cluster is sunlight noise.
[0086] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.
[0087] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.
Claims
1. A method for filtering sunlight noise in open-pit mines using multi-LiDAR fusion, comprising setting up a first LiDAR and a second LiDAR at different installation positions and / or angles, wherein the scanning area of the second LiDAR can cover the area R in which the first LiDAR generates dense sunlight noise, characterized in that... The method for distinguishing between obstacle point clouds and sunlight noise point clouds includes the following steps: Step 1: Obtain the point clouds of the first and second lidars within the region R, and merge them into a dataset; Step 2: Cluster the dataset to obtain N point cloud clusters, and determine whether each point cloud cluster with a first lidar has a point cloud with a second lidar. In scenario one, when the point cloud cluster containing the first lidar contains the point cloud of the second lidar, the point cloud density of the first lidar and the second lidar are compared and matched within the point cloud cluster, and then the similarity between the point clouds of the first lidar and the second lidar is calculated. Scenario 2: When there is no point cloud with a second lidar in the point cloud cluster with the first lidar, calculate the parameters of the point cloud cluster and the parameters of the two beams at the front and rear ends of the emission range of the second lidar located at the center of the point cloud cluster, as well as the proportion of effective point cloud inside the point cloud cluster. Step 3: Based on the calculation results in Step 2, classify the point cloud of the first lidar into obstacle point cloud or sunlight noise point cloud. In scenario one of step two: When the point cloud density of the first lidar is greater than that of the second lidar, the point cloud of the second lidar in the point cloud cluster is circumferentially constructed by adding or subtracting the offset of the line position of the second lidar, and the point cloud of the first lidar located in the line band is extracted to form a new point cloud. or When the point cloud density of the first lidar is less than that of the second lidar, a circumferential beamline is constructed by adding or subtracting an offset from the beamline position of the first lidar within the point cloud cluster. The point cloud of the second lidar located within this beamline is then extracted to form a new point cloud; or When the point cloud density of the first lidar and the second lidar is equal, the point cloud of the first lidar or the second lidar is selected as the new point cloud. The distance between the new point cloud and the corresponding point cloud in the first or second lidar is calculated and used as the similarity score.
2. The method for filtering sunlight noise in open-pit mines using multi-lidar fusion as described in claim 1, characterized in that, Before step two, step one also includes: Delete the ground point cloud from the dataset; Determine whether the dataset is an empty set; The dataset is retained as a non-empty set.
3. The method for filtering sunlight noise in open-pit mines using multi-lidar fusion as described in claim 2, characterized in that, The steps in step three include: S301, determine the similarity and the point cloud similarity threshold. Size; if similarity is less than The new point cloud is the point cloud of the obstacle; if the similarity is greater than 1, then the new point cloud is the point cloud of the obstacle. The new point cloud then contains sunlight noise.
4. The method for filtering sunlight noise in open-pit mines using multi-lidar fusion as described in claim 1, characterized in that, In step two, scenario two, the parameters of the point cloud cluster and the parameters of the two beams at the front and rear ends of the second lidar emission range located at the center of the point cloud cluster are calculated as follows: Construct the bounding box of the point cloud cluster, and obtain the parameters of the bounding box including length, width, height, and horizontal distance between the center point and the second lidar; Calculate the horizontal distances S1 and S2 between the two said wire beams and the second lidar; Calculate the ground spacing between the two wire harnesses, and the maximum height at which an obstacle can be accommodated between the two wire harnesses.
5. A method for filtering sunlight noise in open-pit mines using multi-lidar fusion as described in claim 3, characterized in that, The steps in step three also include: S302, determine if the length and width of the bounding box are less than the ground spacing. If so, the point cloud cluster is a sunlight noise point cloud; otherwise, the point cloud cluster is an obstacle point cloud. Determine whether the sum of the horizontal distance between the center point and the second lidar and the length of the bounding box lies between S1 and S2. If it does, the point cloud cluster is a sunlight noise point cloud; if it does, the point cloud cluster is an obstacle point cloud. Determine whether the height of the bounding box is less than the maximum height that the obstacle can be accommodated between the two line bundles. If it is, then the point cloud of the point cloud cluster is a sunlight noise point cloud. If it is, then the point cloud of the point cloud cluster is an obstacle point cloud.
6. The method for filtering sunlight noise in open-pit mines using multi-lidar fusion as described in claim 5, characterized in that, The step of determining the percentage of effective point clouds within a point cloud cluster includes: Each point in the obstacle point cloud is taken as the origin, and a search area with a radius of r is defined. The search region whose number of internal point clouds exceeds the point count threshold num_threadshold is counted, and the origin corresponding to the search region is taken as a valid point. Calculate the proportion of valid points in each search region relative to all point clouds in the point cloud cluster.
7. A method for filtering sunlight noise in open-pit mines using multi-lidar fusion as described in claim 6, characterized in that, The steps in step three also include: S303, obtain the effective point ratio threshold threadshold and compare it with the ratio of the effective point cloud. If the ratio of the effective point cloud is large, the origin is a sunlight noise point cloud. If the ratio of the effective point cloud is small, the origin is an obstacle point cloud.