Single-line laser glass noise filtering method and device, detection equipment and storage medium

Abstract

The application discloses a single-line laser glass noise point filtering method and device, a detection equipment and a storage medium. The method comprises the following steps: determining target point cloud which may be glass noise points according to the incident angle of points in the point cloud, wherein the target point cloud is composed of points with the incident angle in a set range; and filtering out glass noise points of a set type from the target point cloud. The above technical scheme firstly determines the target point cloud which may be glass noise points, and excludes part of the point cloud which is not glass noise points; and then, according to the type of the glass noise points, the corresponding type of glass noise points is further filtered out from the target point cloud, so that the identification of the glass noise points is more targeted, and the accuracy of the glass noise point filtering is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of point cloud recognition technology, and in particular to a method, apparatus, detection device and storage medium for single-line laser glass noise filtering. Background Technology

[0002] With the modernization of cities, the demands for sanitation services are increasing, and manual cleaning is no longer sufficient for some scenarios. Autonomous vehicles are playing an increasingly important role in the sanitation and cleaning field. Among them, single-line lidar is widely used in the field of low-speed autonomous robots due to its characteristics of directly obtaining environmental depth information, long range, high accuracy, and low cost.

[0003] However, noise often occurs in highly reflective environments such as shopping malls and supermarkets, where glass and marble walls are exposed. Figure 1 This is a schematic diagram of a point cloud. (Example) Figure 1 As shown, in a glass environment, part of the laser beam returns to the glass surface, forming dots (normal dots) on the glass; part of the beam passes through the glass, hits obstacles behind the glass, and then returns; and some dots return in front of the glass. The interference between the emitted and reflected beams causes glass noise. If glass noise cannot be accurately filtered out and is mistakenly identified as obstacles, it will affect the smoothness of robot navigation. Summary of the Invention

[0004] This invention provides a method, apparatus, detection device, and storage medium for single-line laser glass noise filtering, in order to improve the accuracy of glass noise filtering.

[0005] In a first aspect, embodiments of the present invention provide a method for filtering noise on single-line laser glass, comprising:

[0006] The target point cloud that may be glass noise is determined based on the incident angle of the points in the point cloud, wherein the target point cloud is composed of points whose incident angle is within a set range;

[0007] Filter out glass noise of a set type from the target point cloud.

[0008] Secondly, embodiments of the present invention provide a single-line laser glass noise filtering device, comprising:

[0009] The target point cloud determination module is used to determine the target point cloud that may be glass noise based on the incident angle of the points in the point cloud, wherein the target point cloud is composed of points whose incident angle is within a set range;

[0010] A glass noise filtering module is used to filter out glass noise of a set type from the target point cloud.

[0011] Thirdly, embodiments of the present invention provide a detection device, including:

[0012] One or more processors;

[0013] Storage device for storing one or more programs;

[0014] When the one or more programs are executed by the one or more processors, the one or more processors implement the single-line laser glass noise filtering method as described in the first aspect.

[0015] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the single-line laser glass noise filtering method as described in the first aspect.

[0016] This invention provides a method, apparatus, detection device, and storage medium for filtering glass noise using a single-line laser. The method includes: determining a target point cloud that may contain glass noise based on the incident angle of points in a point cloud, wherein the target point cloud consists of points with incident angles within a set range; and filtering out glass noise of a set type from the target point cloud. This technical solution first determines the target point cloud that may contain glass noise, excluding some non-glass noise points; then, according to the type of glass noise, it further filters out the corresponding type of glass noise from the target point cloud, making the identification of glass noise more targeted and improving the accuracy of glass noise filtering. Attached Figure Description

[0017] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0018] Figure 1 This is a schematic diagram of a point cloud.

[0019] Figure 2 This is a flowchart of a single-line laser glass noise filtering method provided in Embodiment 1 of the present invention;

[0020] Figure 3 This is a schematic diagram of an incident angle provided in Embodiment 2 of the present invention;

[0021] Figure 4 This is a flowchart of a single-line laser glass noise filtering method provided in Embodiment 2 of the present invention;

[0022] Figure 5 This is a schematic diagram of a convex curved noise point provided in Embodiment 2 of the present invention;

[0023] Figure 6 This is a schematic diagram of filtering out convex curve-shaped noise points according to Embodiment 2 of the present invention;

[0024] Figure 7 This is a schematic diagram of a method for filtering convex curve-shaped noise based on convex effects and concave characteristics, provided in Embodiment 2 of the present invention.

[0025] Figure 8 This is a schematic diagram of a sawtooth-shaped noise point provided in Embodiment 2 of the present invention;

[0026] Figure 9 This is a schematic diagram of a sawtooth point cloud provided in Embodiment 2 of the present invention;

[0027] Figure 10 This is a schematic diagram of a cluster of flickering noise points provided in Embodiment 2 of the present invention;

[0028] Figure 11A This is a schematic diagram of constructing a spacetime graph according to Embodiment 2 of the present invention;

[0029] Figure 11B This is a schematic diagram of edge extraction provided in Embodiment 2 of the present invention;

[0030] Figure 11C This is a schematic diagram of edge extraction provided in Embodiment 2 of the present invention;

[0031] Figure 12 This is a schematic diagram illustrating a filtering implementation based on time-domain tracking results provided in Embodiment 2 of the present invention;

[0032] Figure 13 This is a schematic diagram of pruning a target according to Embodiment 2 of the present invention;

[0033] Figure 14 This is a schematic diagram illustrating the implementation of target pruning according to Embodiment 2 of the present invention;

[0034] Figure 15 This is a schematic diagram of the structure of a single-line laser glass noise filtering device provided in Embodiment 3 of the present invention;

[0035] Figure 16 This is a schematic diagram of the hardware structure of a detection device provided in Embodiment 4 of the present invention. Detailed Implementation

[0036] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, the embodiments and features described herein can be combined with each other unless otherwise specified. It should also be noted that, for ease of description, only the parts relevant to the present invention are shown in the drawings, not the entire structure.

[0037] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processes, many of these steps can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the steps can be rearranged. The process can be terminated when its operation is complete, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0038] It should be noted that the concepts of "first" and "second" mentioned in the embodiments of the present invention are only used to distinguish different devices, modules, units or other objects, and are not used to limit the order of functions performed by these devices, modules, units or other objects or their interdependencies.

[0039] Example 1

[0040] Figure 2 This is a flowchart of a single-line laser glass noise filtering method provided in Embodiment 1 of the present invention. This embodiment is applicable to filtering glass noise from point clouds measured by a single-line laser. Specifically, this single-line laser glass noise filtering method can be executed by a single-line laser glass noise filtering device, which can be implemented by software and / or hardware and integrated into the detection device. Furthermore, the detection device is mobile, for example, it can be a vehicle. The detection device is equipped with a single-line lidar, and the laser source emits a single-line beam, which reacts faster in terms of angular frequency and sensitivity, and is more accurate in terms of distance and precision in testing obstacles. This embodiment provides a system structure plus time domain method that minimizes the risk of false filtering while filtering glass noise.

[0041] like Figure 3 As shown, the method specifically includes the following steps:

[0042] S110. Determine a target point cloud that may be glass noise based on the incident angle of the points in the point cloud, wherein the target point cloud is composed of points whose incident angle is within a set range.

[0043] Specifically, glass noise typically occurs when the incident angle is approximately 90 degrees. In environments where glass wall lines can be extracted, the incident angle of a point can be obtained in real time based on the position information of the point in the point cloud illuminating the environment, thus filtering out potential glass noise targets. In scenarios where glass wall lines cannot be actively extracted, such as when the detection equipment is not equipped with ultrasonic radar or lacks the sensor equipment or related software functions for extracting wall lines, the scenario where the detection equipment moves along the glass wall can be considered.

[0044] In this embodiment, the angle between a point in the point cloud and the X-axis is used as the incident angle. Based on the collected point cloud, the incident angle of laser irradiation at various locations on the glass is determined. The number of points in the point cloud corresponds to the number of locations irradiated, and thus the number of incident angles. Points within a defined range around an incident angle of 90° constitute the target point cloud. Subsequent filtering operations target the target point cloud, thereby eliminating some non-glass noise points.

[0045] In one embodiment, before determining the target point cloud that may be glass noise based on the incident angle of the points in the point cloud, the method further includes: establishing a Cartesian coordinate system; and taking the angle between each point in the point cloud and the X-axis direction of the Cartesian coordinate system as the incident angle of the corresponding point.

[0046] Figure 3 This is a schematic diagram of an incident angle provided in Embodiment 2 of the present invention. Figure 3 As shown, the point cloud coordinates are established in a Cartesian coordinate system centered on the lidar (specifically, a front-left-sky coordinate system, i.e., front-x, left-y, sky-right). The point cloud is an ordered point cloud. Figure 3 In this context, -120 degrees is the starting point and 120 degrees is the ending point. Different lidars may have different resolutions. For example, if the lidar resolution is 0.5 degrees, then 481 points can be collected in each frame. Figure 3 The black-filled area in the image refers to the range of the target point cloud formed by points whose angle with the x-axis direction is within a set range (around 90 degrees).

[0047] S120. Filter out glass noise of a set type from the target point cloud.

[0048] In this embodiment, glass noise of the corresponding type is further filtered out from the target point cloud according to the type of glass noise. For example, according to structural morphology, glass noise can be divided into four types: convex curves, jagged edges, occasional flashing cluster noise, and clustered flashing noise that appears for a long time at the same location. Among them, for convex curves, filtering can be performed based on their structural characteristics, such as using multi-segment fitting and based on the slope and intercept information of the line segments; for jagged clustered noise, filtering can also be performed based on their structural characteristics, such as the proportion of sharp angles in the cluster; for occasional or long-term flashing cluster noise, since it is difficult to extract effective features from the structure, a temporal tracking method can be used for filtering.

[0049] The present invention provides a single-line laser glass noise filtering method. First, the target point cloud that may be glass noise is identified, and some non-glass noise point clouds are excluded. Then, according to the type of glass noise, the corresponding type of glass noise is further filtered out from the target point cloud, making the identification of glass noise more targeted. For different types of glass noise, their structural features or temporal features can be fully utilized for filtering, which improves the accuracy of glass noise filtering and avoids the impact of glass noise on the navigation of the detection equipment.

[0050] Example 2

[0051] Figure 4 This is a flowchart illustrating a single-line laser glass noise filtering method according to Embodiment 2 of the present invention. This embodiment is an optimization based on the above embodiments, and provides a detailed description of single-line laser glass noise filtering. It should be noted that technical details not described in detail in this embodiment can be found in any of the above embodiments.

[0052] Specifically, such as Figure 4 As shown, the method specifically includes the following steps:

[0053] S210. Determine a target point cloud that may be glass noise based on the incident angle of the points in the point cloud, wherein the target point cloud is composed of points whose incident angle is within a set range.

[0054] S220. Clustering is performed based on the distance difference between the points in the target point cloud and the origin, as well as the point numbers.

[0055] In this embodiment, for the pre-screened target point cloud suspected of being glass noise, preliminary clustering can be performed using features such as differences in distance (Range) and index between points. The point cloud data of a single-line LiDAR is a series of ordered point clouds. The Index refers to the point's sequence number, and the Range refers to the distance from the point to the LiDAR origin. Using information such as distance and sequence number, the target point cloud can be clustered to obtain multiple clustering results. Then, for each clustering result, glass noise of a specified type is filtered out, improving the accuracy of identification and reducing the risk of false filtering.

[0056] S230. Perform multi-segment fitting on the points in each clustering result, and filter out the noise points in the form of convex curves based on the slope and intercept information of the fitted line segments.

[0057] Among them, the glass noise type setting includes convex curve noise.

[0058] Figure 5 This is a schematic diagram of a convex curved noise point provided in Embodiment 2 of the present invention. Figure 5 As shown, convex curved noise points are morphologically oriented towards the origin of the lidar and form a 90-degree angle with the positive X-axis. To identify this type of convex curved noise, multi-segment fitting can be performed sequentially for each cluster. For the fitted straight lines, the line segments formed by glass noise are filtered out based on the slope and intercept characteristics. The points in these line segments are the glass noise points.

[0059] Figure 6 This is a schematic diagram illustrating a method for filtering convex curve-shaped noise points according to Embodiment 2 of the present invention. Figure 6 As shown, the leftmost image represents the original point cloud morphology; the middle image shows the multi-segment fitting effect, where two line segments were fitted: segment 0 and segment 1; the right image illustrates one method of multi-segment fitting: split-line fitting. For segment 0 and segment 1, their slope and intercept can be calculated respectively. If the slope and intercept of a line segment conform to the convex curve noise morphology (the slope and intercept belong to the slope range and intercept range of the convex curve noise morphology, respectively), then this line segment points towards the origin of the lidar and forms a 90-degree angle with the positive X-axis. The points in this line segment represent the convex curve noise morphology.

[0060] In one embodiment, the method further includes removing corner segments and rectangular segments based on the convexity and concaveness characteristics of continuous line segments in the clustering results.

[0061] Specifically, considering that glass noise points are line segments extending outward from the wall, the non-glass noise point line segments corresponding to the corners of the walls and the rectangular shapes (such as boxes or other obstacles with rectangular outlines) in the clustering results can be removed in advance based on the convex and concave characteristics of continuous line segments, so as to avoid misidentifying the corners or rectangular shapes as glass noise points.

[0062] Figure 7 This is a schematic diagram illustrating how, according to Embodiment 2 of the present invention, noise in the form of convex curves is filtered based on convexity effects and concave characteristics. Figure 7 As shown, in the section on the left where the fitted line segments are: (a) the fitted line segments satisfy the slope range and intercept range of the convex curve noise, and the corresponding points can be identified as convex curve noise and thus filtered out; (b) the fitted line segments do not satisfy the slope range and intercept range of the convex curve noise, and the corresponding points will not be identified as convex curve noise. In the section on the right where non-glass noise is excluded based on the convex effect and concave characteristics: (c) represents convex line segments, which do not correspond to the shape of glass noise, but to real obstacles in the environment, so the corresponding points will not be identified as convex curve noise; (d) represents concave line segments, which conform to the shape of glass noise extending from the glass wall, so the corresponding points can be identified as convex curve noise and thus filtered out.

[0063] In one embodiment, the method further includes: retaining line segments whose slope and intercept information obtained by cumulative fitting within a set time period does not conform to the convex curve noise feature; during the processing of the point cloud of the current frame, matching the line segments in the current frame whose slope and intercept information does not conform to the convex curve noise feature with the retained line segments; if a match is found, identifying the points in the line segments in the current frame whose slope and intercept information does not conform to the convex curve noise feature as real obstacles.

[0064] Specifically, for some flat objects, the point cloud morphology on their surface can exhibit line segment shapes similar to convex curves at certain angles between the detection device and the obstacle. To avoid identifying such point clouds as glass noise, considering that glass noise moves with the movement of the detection device, while static flat objects exist stably in a specific position, the point cloud morphology of glass noise varies depending on the relative position and orientation when the detection device moves. Therefore, during frame-by-frame processing, line segments whose slope and intercept information cumulatively detected within a set time period (across multiple historical frames) do not conform to the convex curve noise characteristics are retained (they can be identified as stable, real obstacles and therefore will not be filtered). When processing the latest current frame point cloud, line segments that do not conform to the convex curve noise characteristics are matched with the cumulatively detected line segments based on their position, length, direction, and other characteristics. If a match is successful, it indicates that the line segments in the current frame that do not conform to the convex curve noise characteristics are also stable, real obstacles, thus avoiding false filtering of real obstacles and improving the accuracy of glass noise filtering.

[0065] S240. Filter out jagged noise points based on the proportion of points in the jagged point cloud in each clustering result.

[0066] Among them, the glass noise type is defined as jagged noise.

[0067] Figure 8 This is a schematic diagram of a sawtooth-shaped noise point provided in Embodiment 2 of the present invention. Figure 8 As shown, some glass noise points exhibit the characteristic of connecting continuous points with the line between the current point and the origin of the lidar, which is approximately 0 degrees or 180 degrees apart. Furthermore, the distance between continuous points differs from the distance pattern of the point cloud formed on the surface of the actual obstacle. Based on these two characteristics, jagged noise points are filtered out.

[0068] In one embodiment, filtering out the jagged noise points based on the proportion of points in the jagged point cloud in each clustering result includes: in each clustering result, if the angle between the line connecting a point and its adjacent points and the line connecting the point and the origin is within a specified range of 0° or 180°, then the point is a point in the jagged point cloud (these consecutive points or adjacent points constitute the jagged point cloud); if the proportion of points in the jagged point cloud in the corresponding clustering result exceeds a set threshold, then the points in the jagged point cloud are determined as jagged noise points.

[0069] Figure 9 This is a schematic diagram of a sawtooth point cloud provided in Embodiment 2 of the present invention. Figure 9As shown, points 1, 2, and 3 form a cluster with corresponding angles θ1, θ2, and θ3, respectively. θ1 is close to 0 degrees, and θ2 is close to 180 degrees. Therefore, this cluster contains two jagged point clouds. Based on this, we can determine the percentage of points in the jagged point cloud within a cluster relative to the total number of points in that cluster. If this percentage exceeds a set threshold, the points in the jagged point cloud can be identified as jagged noise. The threshold can be set based on statistical results and actual needs.

[0070] S250. Perform temporal tracking on the points in each clustering result, and filter out the first type of clustered flashing noise based on the tracking results.

[0071] Among them, the glass noise of the set type includes the first type of clustered flickering noise, and the occurrence frequency of the first type of clustered flickering noise is lower than the threshold (which can be understood as occasional clustered flickering noise).

[0072] Figure 10 This is a schematic diagram of a cluster of flickering noise points provided in Embodiment 2 of the present invention. Figure 10 As shown in this embodiment, after structural filtering, there are still some point clouds with clusters of flashing noise that do not have obvious structural features. They cannot be distinguished from real obstacles in terms of structural features. For these points, a spatiotemporal graph can be constructed to track the temporal and spatial structure of point clouds in multiple consecutive frames.

[0073] In one embodiment, temporal tracking is performed on points in each clustering result, and the first type of clustered flickering noise is filtered out based on the tracking results, including:

[0074] Edges are extracted based on the position of points in each clustering result across multiple consecutive frames;

[0075] The target (obj) is constructed based on edge connections;

[0076] Prune the target;

[0077] Based on the pruning results, the first type of clustered flashing noise points and real obstacle points were obtained by filtering based on the number of layers at the edge.

[0078] Figure 11A This is a schematic diagram illustrating the construction of a spacetime graph according to Embodiment 2 of the present invention. Figure 11A As shown, an 8-frame spatiotemporal graph can be constructed using the current frame and historical frames (e.g., 7 frames). The collected 8 consecutive frames of point cloud data are stacked in the same three-dimensional space. For example, the 8 consecutive frames of point cloud data can be constructed simultaneously in the same three-dimensional space in the form of 8 layers of feature points, where each frame of point cloud data corresponds to one layer of feature points.

[0079] Figure 11BThis is a schematic diagram illustrating edge extraction according to Embodiment 2 of the present invention. Figure 11B As shown, an edge is built by accumulating 8 consecutive frames of point cloud data. Specifically, feature points in the 3D space are connected point-to-point to create multiple edge lines corresponding to the target. During the edge building process, the connection between adjacent feature points in the same frame (lateral connection) is determined by Euclidean distance (if the Euclidean distance meets the first preset range, a lateral connection is performed); the connection between a feature point in a frame and a historical frame (vertical connection) is determined by searching upwards within two frames, calculating the Euclidean distance between the historical frame point cloud (after position prediction) and the current frame point cloud (if the Euclidean distance meets the second preset range, a vertical connection is performed).

[0080] Figure 11C This is a schematic diagram illustrating edge extraction according to Embodiment 2 of the present invention. Figure 11C As shown, an object is constructed by connecting edges. Once all edges are established, the object can be divided into individual objects based on the connections between the points.

[0081] This embodiment first prunes the object (obj), and then roughly filters out noise points and real obstacle points based on the edge layer number (age). Considering that only considering the age of the object might misidentify moving obstacle points as first-type clumped flashing noise points, the dynamic object (obj) after tracking can also be corrected to distinguish between moving obstacles and first-type clumped flashing noise points.

[0082] Figure 12 This is a schematic diagram illustrating an implementation of filtering based on time-domain tracking results according to Embodiment 2 of the present invention. Figure 12 As shown, after filtering based on structure, for clustered flickering noise, noise and real obstacles can be distinguished based on the edge layer number. Furthermore, the state of the points (dynamic, static, or unstable) can be used to correct for noise and real obstacles. Specifically, if a noise point corresponds to a static or unstable point, it can be ultimately considered a type I clustered flickering noise; if a noise point corresponds to a dynamic point, it can be ultimately considered a dynamic point of a dynamic obstacle; and points of real obstacles can be directly identified as dynamic, static, or unstable points.

[0083] In one embodiment, pruning the target includes: obtaining a confidence level based on the number of edge layers, the number of edges in each layer, the number of points, the edge length and the continuity of the direction, and the dynamic and static properties of the target; and pruning the target based on the confidence level.

[0084] In this embodiment, considering that the establishment of edge connections between the first type of clustered flashing noise and adjacent real objects can easily lead to missed filtering, it is necessary to prune each object first. A confidence level is obtained based on the object's layer number, the number of edges per layer, the number of points, edge length and directional continuity, and the object's dynamic and static properties, serving as the basis for pruning. For example, vertical connections with a confidence level higher than this level can be considered edge connections of real objects and can be retained; vertical connections with a confidence level lower than this level may be edge connections established between the first type of clustered flashing noise and adjacent real objects, and these edge connections need to be pruned.

[0085] Figure 13 This is a schematic diagram illustrating a target pruning method according to Embodiment 2 of the present invention. Figure 13 As shown, the connections within the elliptical region represent the top and bottom connections to be cut off.

[0086] Figure 14 This is a schematic diagram illustrating the implementation of target pruning according to Embodiment 2 of the present invention. Figure 14 As shown, the state of an object (including static, dynamic, and uncertain, also known as unstable objects) can be distinguished based on the shape of its edges. A feature, the projection length of the connected edges on the xy-plane, is calculated to obtain the thresholds for static and unstable targets. Dynamic objects are not pruned. For static objects, if the projection length of the connected edges on the xy-plane is greater than the corresponding threshold (0.05), the connections are pruned. For unstable objects, if the projection length of the connected edges on the xy-plane is greater than the corresponding threshold (0.2), the connections are pruned. Based on the pruning results, noise and real obstacle points are obtained by edge-based age filtering. Furthermore, the object state can be used to correct noise and real obstacles, resulting in the first type of clumped flashing noise and real obstacle points.

[0087] S260. Filter out the second type of clumped glass noise based on the target's location, point cloud morphology, and number of consecutive tracking iterations in each clustering result.

[0088] The glass noise type is defined as including the second type of clustered flickering noise, where the frequency of occurrence of the second type of clustered flickering noise is higher than or equal to a threshold (which can be understood as clustered flickering noise appearing at the same location for a long time). In this embodiment, the second type of clustered glass noise mainly refers to the glass noise detected by the detection device in a stationary state, and can be determined comprehensively based on the target's position (e.g., the target's position in the lidar coordinate system, specifically the positional relationship between the target and the lidar coordinate system in the Y direction), point cloud morphology (e.g., the size of sharp angles in the target and the number of points in the target), and the number of consecutive tracking times (the number of times the target is detected within a certain period of time).

[0089] In one embodiment, filtering out the second type of clumped glass noise from the target's location, point cloud morphology, and number of consecutive tracking iterations in each clustering result includes:

[0090] Filter the targets in each clustering result according to the following criteria:

[0091] The target state includes moving, static, or uncertain. The target state can be determined based on the number of consecutive tracking. The method for determining the target state based on multiple tracking of the target can be referred to the above process for distinguishing the state of clustered flashing noise points.

[0092] The target is within ±15 degrees of the Y direction in the lidar coordinate system and the range of variation is within the set range (i.e., filtering is performed based on the target's position);

[0093] The target contains sharp angles (i.e., filtering is performed based on the point cloud shape of the target), for example, the angle of the sharp angle is less than the set angle threshold;

[0094] The number of points in the target is within a set range (i.e., filtered according to the point cloud shape of the target);

[0095] A movement speed within a set speed range around 0 indicates that the detection device is stationary, thus providing a prerequisite for filtering out second-type clustered flashing noise.

[0096] Points that meet all the above conditions are classified as second-type clustered flickering noise points.

[0097] The second embodiment of this invention provides a single-line laser glass noise filtering method, which optimizes the above embodiment. First, it identifies the target point cloud that may contain glass noise and excludes some non-glass noise point clouds. Then, it further filters out the corresponding type of glass noise from the target point cloud according to the type of glass noise, making the glass noise identification more targeted. For different types of glass noise, its structural features or temporal features can be fully utilized for filtering. Furthermore, information such as distance and sequence number can be used to cluster the target point cloud, obtaining multiple clustering results. Then, for each clustering result, a set type of glass noise is filtered out. To improve the accuracy of identification, based on the convex and concave characteristics of continuous line segments, line segments corresponding to corners and non-glass noise points corresponding to rectangular shapes in the clustering results are removed, avoiding misidentification of corners or rectangular points as glass noise. By matching line segments in the current frame whose slope and intercept information do not conform to the convex curve noise characteristics with the retained line segments, the false filtering of real obstacles is avoided, improving the accuracy of glass noise filtering. In addition to structural filtering, operations such as constructing spatiotemporal maps, pruning, and dynamic correction are also used to filter clustered flashing noise, improving the comprehensiveness of glass noise filtering.

[0098] Example 3

[0099] Figure 15 This is a schematic diagram of a single-line laser glass noise filtering device provided in Embodiment 3 of the present invention. Figure 15 As shown, the single-line laser glass noise filtering device provided in this embodiment includes:

[0100] The target point cloud determination module 310 is used to determine a target point cloud that may be glass noise based on the incident angle of the points in the point cloud, wherein the target point cloud is composed of points whose incident angle is within a set range;

[0101] The glass noise filtering module 320 is used to filter out glass noise of a set type from the target point cloud.

[0102] The single-line laser glass noise filtering device provided in Embodiment 3 of the present invention first determines the target point cloud that may be glass noise and excludes some non-glass noise point clouds; then, according to the type of glass noise, it further filters out the corresponding type of glass noise from the target point cloud, making the identification of glass noise more targeted and improving the accuracy of glass noise identification.

[0103] Based on the above embodiments, the device further includes:

[0104] The system establishment module is used to establish a Cartesian coordinate system before determining the target point cloud that may be glass noise based on the incident angle of the points in the point cloud;

[0105] The incident angle determination module is used to take the angle between each point in the point cloud and the X-axis direction of the Cartesian coordinate system as the incident angle of the corresponding point.

[0106] Based on the above embodiments, the glass noise filtering module 320 includes:

[0107] Clustering unit, used to cluster points in the target point cloud based on the distance difference between the points and the origin and the point's index;

[0108] The filtering unit is used to filter out glass noise of a specified type from each clustering result.

[0109] Based on the above embodiments, the set type of glass noise includes convex curved noise;

[0110] The filter unit is specifically used for:

[0111] For each clustering result, a multi-line segment fit is performed on the points, and the noise points in the form of the convex curve are filtered out based on the slope and intercept information of the fitted line segments.

[0112] Based on the above embodiments, the device further includes:

[0113] The removal module is used to remove corner segments and rectangular segments based on the convex and concave characteristics of continuous line segments in the clustering results.

[0114] Based on the above embodiments, it also includes:

[0115] Line segments whose slope and intercept information obtained by cumulative fitting within a set time period do not conform to the noise characteristics of the convex curve form will be retained.

[0116] During the processing of the point cloud of the current frame, line segments in the current frame whose slope and intercept information do not conform to the convex curve form noise feature are matched with the retained line segments.

[0117] If a match is found, points in line segments in the current frame whose slope and intercept information do not conform to the convex curve noise feature are identified as real obstacles.

[0118] Based on the above embodiments, the glass noise type is defined to include jagged noise;

[0119] The filter unit is specifically used for:

[0120] The jagged noise points are filtered out based on the proportion of points in the jagged point cloud in each clustering result.

[0121] Based on the above embodiments, the jagged noise points are filtered out according to the proportion of points in the jagged point cloud in each clustering result, including:

[0122] In each clustering result, if the angle between the line connecting a point to its neighboring points and the line connecting the point to the origin is within a specified range of 0° or 180°, then the point is a point in the jagged point cloud.

[0123] If the proportion of points in the jagged point cloud to the corresponding clustering results exceeds a set threshold, then the points in the jagged point cloud are identified as jagged noise points.

[0124] Based on the above embodiments, the set type of glass noise includes a first type of clustered flashing noise, and the occurrence frequency of the first type of clustered flashing noise is lower than a threshold.

[0125] The filter unit is specifically used for:

[0126] Temporal tracking is performed on the points in each clustering result, and the first type of clustered flashing noise points are filtered out based on the tracking results.

[0127] Based on the above embodiments, the step of performing temporal tracking on points in each clustering result and filtering out the first type of clustered flickering noise points according to the tracking results includes:

[0128] Edges are extracted based on the position of points in each clustering result across multiple consecutive frames.

[0129] The target is formed based on the edge connections described above;

[0130] Prune the target;

[0131] Based on the pruning results, the first type of clustered flashing noise points and real obstacle points are obtained by filtering based on the number of layers of the edge.

[0132] Based on the above embodiments, pruning of the target includes:

[0133] The confidence level is obtained based on the number of edge layers, the number of edges in each layer, the number of points, the edge length, the continuity of direction, and the dynamic and static properties of the target.

[0134] The target is pruned based on the confidence level.

[0135] Based on the above embodiments, the set type of glass noise includes a second type of clustered flashing noise, and the occurrence frequency of the second type of clustered flashing noise is higher than or equal to a threshold.

[0136] The filter unit is specifically used for:

[0137] The second type of clumped glass noise is filtered out based on the target's location, point cloud morphology, and number of consecutive tracking iterations in each clustering result.

[0138] Based on the above embodiments, the second type of clustered glass noise is filtered out according to the target's location, point cloud morphology, and number of consecutive tracking iterations in each clustering result, including:

[0139] Filter the targets in each clustering result according to the following criteria:

[0140] The target state includes moving, static, or uncertain states;

[0141] The target is within ±15 degrees of the Y direction in the lidar coordinate system and the range of variation is within the set range;

[0142] The target contains a sharp angle;

[0143] The number of target midpoints is within the set quantity range;

[0144] Movement speed is within the set speed range around 0.

[0145] The single-line laser glass noise filtering device provided in Embodiment 3 of the present invention can be used to perform the single-line laser glass noise filtering method provided in any of the above embodiments, and has the corresponding functions and beneficial effects.

[0146] Example 4

[0147] Figure 16 This is a schematic diagram of the hardware structure of a detection device according to Embodiment 4 of the present invention. The detection device is movable. Figure 16 As shown, the detection device provided in this application includes a single-line lidar 400, a storage device 420, a processor 410, and a computer program stored on the storage device and executable on the processor. When the processor 410 executes the program, it implements the above-described single-line lidar glass noise filtering method.

[0148] The detection device may also include a storage device 420; the processor 410 in the detection device may be one or more. Figure 16 Taking a processor 410 as an example; storage device 420 is used to store one or more programs; the one or more programs are executed by the one or more processors 410, so that the one or more processors 410 implement the single-line laser glass noise filtering method as described in the embodiments of this application.

[0149] The detection equipment also includes: a communication device 430, an input device 440, and an output device 450.

[0150] The processor 410, storage device 420, communication device 430, input device 440, and output device 450 in the detection device can be connected via a bus or other means. Figure 16 Taking the example of a connection between China and Israel via a bus.

[0151] Input device 440 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the detection device. Output device 450 may include display devices such as a display screen.

[0152] The communication device 430 may include a receiver and a transmitter. The communication device 430 is configured to perform information transmission and reception communication under the control of the processor 410.

[0153] Storage device 420, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the single-line laser glass noise filtering method described in the embodiments of this application (e.g., the target point cloud determination module 310 and the glass noise filtering module 320 in the single-line laser glass noise filtering device). Storage device 420 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the detection device, etc. In addition, storage device 420 may include a high-speed random access storage device, and may also include a non-volatile storage device, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, storage device 420 may further include storage devices remotely located relative to processor 410, which can be connected to the detection device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0154] Based on the above embodiments, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a single-line laser glass noise filtering device, implements the single-line laser glass noise filtering method in any of the above embodiments of the present invention. The method includes: determining a target point cloud that may be glass noise based on the incident angle of points in the point cloud, wherein the target point cloud is composed of points with incident angles within a set range; and filtering out glass noise of a set type from the target point cloud.

[0155] The storage medium containing computer-executable instructions provided in this embodiment of the invention can be any combination of one or more computer-readable media, such as computer-readable signal media or storage media. Computer-readable storage media can be, for example, but not limited to: electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable CD-ROMs, optical storage devices, magnetic storage devices, or any suitable combination thereof. A computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0156] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device.

[0157] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency (RF), etc., or any suitable combination thereof.

[0158] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0159] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0160] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A single-line laser glass noise filtering method, characterized in that, The method comprises the following steps: determining a target point cloud which may be glass noise points according to the incidence angles of points in a point cloud, wherein the target point cloud is composed of points with incidence angles within a set range; filtering out glass noise points of a set type from the target point cloud; the step of filtering out glass noise points of a set type from the target point cloud comprises the following steps: clustering according to the distance differences between points in the target point cloud and the origin and the serial numbers of the points; filtering out glass noise points of a set type from each clustering result; the glass noise points of the set type comprise convex curve noise points; the step of filtering out glass noise points of a set type from each clustering result comprises the following steps: performing multi-segment fitting on the points in each clustering result, and filtering out convex curve noise points according to the slope and intercept information of the fitted line segments.

2. The method of claim 1, wherein, Before the step of determining a target point cloud which may be glass noise points according to the incidence angles of points in a point cloud, the method further comprises the following steps: establishing a Cartesian coordinate system; taking the included angle between each point in the point cloud and the X-axis direction of the Cartesian coordinate system as the incidence angle of the corresponding point.

3. The method of claim 1, wherein, The method further comprises the following steps: removing corner line segments and rectangular-shaped line segments according to the convexity and concavity of continuous line segments in the clustering results.

4. The method of claim 1, wherein, The method further comprises the following steps: retaining line segments with slope and intercept information accumulated within a set time period which do not meet the convex curve noise point characteristics; in the process of processing the point cloud of a current frame, matching the line segments with slope and intercept information which do not meet the convex curve noise point characteristics in the current frame with the retained line segments; if the matching is successful, identifying the points in the line segments with slope and intercept information which do not meet the convex curve noise point characteristics in the current frame as real obstacles.

5. The method of claim 1, wherein, The glass noise points of the set type comprise sawtooth noise points; the step of filtering out glass noise points of a set type from each clustering result comprises the following steps: filtering out the sawtooth noise points according to the proportion of points in a sawtooth point cloud in each clustering result.

6. The method of claim 5, wherein, The step of filtering out the sawtooth noise points according to the proportion of points in a sawtooth point cloud in each clustering result comprises the following steps: in each clustering result, if the angle between the line connecting a point and its adjacent points and the line connecting the point and the origin is within a specified range around 0° or 180°, the point is a point in a sawtooth point cloud; if the proportion of points in the sawtooth point cloud in the corresponding clustering result exceeds a set threshold, the points in the sawtooth point cloud are determined as the sawtooth noise points.

7. The method of claim 1, wherein, The glass noise points of the set type comprise first type cluster flicker noise points, and the occurrence frequency of the first type cluster flicker noise points is lower than a threshold value; the step of filtering out glass noise points of a set type from each clustering result comprises the following steps: performing time domain tracking on the points in each clustering result, and filtering out the first type cluster flicker noise points according to the tracking results.

8. The method of claim 7, wherein, The step of performing time domain tracking on the points in each clustering result and filtering out the first type cluster flicker noise points according to the tracking results comprises the following steps: extracting edges according to the positions of the points in each clustering result in continuous multiple frames; connecting the edges to form a target; pruning the target; based on the number of layers of the edges, screening the first type cluster flicker noise points and real obstacle points according to the pruning results.

9. The method of claim 8, wherein, The pruning the target comprises: obtaining a confidence degree according to the number of layers of the edge, the number of edges of each layer, the number of points, the continuity of the edge length and direction, and the dynamic and static properties of the target; pruning the target according to the confidence degree.

10. The method of claim 1, wherein, The set type of glass noise points includes second type of group flickering noise points, and the occurrence frequency of the second type of group flickering noise points is higher than or equal to a threshold value; The filtering of the set type of glass noise points from each clustering result comprises: filtering the second type of group glass noise points according to the position of the target in each clustering result, the point cloud shape, and the number of continuous tracking times.

11. The method of claim 10, wherein, The filtering of the second type of group glass noise points according to the position of the target in each clustering result, the point cloud shape, and the number of continuous tracking times comprises: filtering the target in each clustering result according to the following conditions: a target state, the target state including moving, static or uncertain; the target being within a range of 15 degrees in the Y direction of the laser radar coordinate system and a set range of variation range; a sharp angle existing in the target; the number of points in the target being within a set number range; the moving speed being within a set speed range around 0.

12. A single-line laser glass noise filter apparatus, characterized by, The method comprises: a target point cloud determination module configured to determine target point clouds that are likely to be glass noise points according to incident angles of points in a point cloud, wherein the target point cloud is composed of points with incident angles within a set range; a glass noise point filtering module configured to filter set types of glass noise points from the target point cloud; The glass noise point filtering module comprises: a clustering unit configured to cluster points in the target point cloud according to distance differences between the points and an origin point and serial numbers of the points; a filtering unit configured to filter set types of glass noise points from each clustering result; The set type of glass noise points includes convex curve type noise points; The filtering unit is specifically configured to: perform multi-line segment fitting on points in each clustering result, and filter the convex curve type noise points according to slope and intercept information of the fitted line segments.

13. A detection device, characterized by The method comprises: a single-line laser radar; one or more processors; a storage device configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the single-line laser glass noise point filtering method according to any one of claims 1-11.

14. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the single-line laser glass noise point filtering method according to any one of claims 1-11.

Images