Environment perception method based on lidar and high-precision map

Abstract

The present application relates to the technical field of autonomous driving, and relates to an environment perception method based on lidar and a high-precision map. The method comprises: acquiring environment point cloud data in real time via lidar; on the basis of a high-precision map, delineating a region of interest; performing object recognition and classification on the environment point cloud data within the region of interest; classifying objects within the region of interest into a dynamic obstacle layer or a static obstacle layer on the basis of motion attributes, geometric characteristics, and point cloud density of the objects; and on the basis of road attribute information of the high-precision map, optimizing classification and tracking results of the objects, and assisting in path planning and decision-making. The present application can improve the environment perception capability of a vehicle during driving, effectively reduce computational burden, reduce the false detection rate, and improve the adaptability and decision reliability of the system in complex scenarios.

Description

An environmental perception method based on lidar and high-precision maps Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to an environmental perception method based on LiDAR and high-precision maps. Background Technology

[0002] Currently, environmental perception technology is one of the core modules for achieving safe and reliable operation in autonomous driving systems. Existing environmental perception methods typically rely on LiDAR and image sensors to provide dynamic and static target information through target detection and classification. However, limited by the way environmental point cloud data is processed, existing technologies often suffer from low classification accuracy and high computational cost in complex scenarios. For example, when faced with a mixed distribution of dynamic and static targets or irregularly shaped objects, existing methods struggle to effectively distinguish target categories, leading to deviations in path planning and obstacle avoidance decisions. Furthermore, redundant data processing increases the system's computational burden, further impacting real-time performance. Therefore, there is an urgent need for an improved method that can enhance the accuracy of environmental perception and optimize computational efficiency to meet the complex requirements of autonomous driving. Summary of the Invention

[0003] Based on this, an environment perception method based on lidar and high-precision maps is provided, the method comprising:

[0004] Real-time acquisition of environmental point cloud data via lidar;

[0005] Regions of interest are defined based on high-precision maps;

[0006] Target identification and classification are performed on the environmental point cloud data within the region of interest;

[0007] Based on the target's motion attributes, geometric characteristics, and point cloud density, the target within the region of interest is divided into a dynamic obstacle layer or a static obstacle layer.

[0008] The classification and tracking results of targets are optimized based on the road attribute information of the high-precision map, and path planning and decision-making are assisted.

[0009] As an optional implementation, the method of dividing the region of interest based on a high-precision map includes:

[0010] Read the global road boundary information from the high-precision map;

[0011] Based on the vehicle's location, obtain local road boundary information within a preset range near the vehicle;

[0012] Construct the point cloud data boundary corresponding to the local road boundary;

[0013] The region within the local road boundary is defined as the region of interest.

[0014] As an optional implementation, the method further includes:

[0015] The area outside the local road boundary is defined as the region of non-interest.

[0016] The targets within the non-interest area are assigned to the static obstacle layer.

[0017] As an optional implementation, the step of performing target identification and classification on the environmental point cloud data within the region of interest includes:

[0018] Preliminary target type identification based on target motion characteristics and geometric features;

[0019] The preliminary identification results are verified and optimized based on the regional attributes of the high-precision map.

[0020] As an optional implementation, the step of dividing the target within the region of interest into a dynamic obstacle layer or a static obstacle layer based on the target's motion attributes, geometric characteristics, and point cloud density includes:

[0021] The targets with dynamic motion attributes are assigned to the dynamic obstacle layer;

[0022] The target with static motion attributes is assigned to the static obstacle layer;

[0023] Small-sized targets with point cloud density below a preset threshold are assigned to the static obstacle layer;

[0024] The abnormal point cloud targets that failed to be detected are assigned to the static obstacle layer.

[0025] As an optional implementation, the method further includes:

[0026] The original point cloud shape of the target is preserved in the static obstacle layer.

[0027] As an optional implementation, the road attribute information of the high-precision map includes slope attributes, straight road attributes, and sidewalk attributes.

[0028] As an optional implementation, the optimization of target classification and tracking results based on the road attribute information of the high-precision map includes:

[0029] Based on the pedestrian walkway attributes, the speed of pedestrian targets is verified and limited;

[0030] Based on the aforementioned ramp attributes, the ramp surface is identified and optimized;

[0031] Based on the straight-line attribute, the target direction information is verified and the tracking trajectory is optimized.

[0032] As an optional implementation, the identification and optimization of the ramp surface based on the ramp attributes includes:

[0033] Obtain the slope angle of the ramp;

[0034] The ground segmentation height threshold is dynamically set according to the tilt angle;

[0035] The slope road surface is segmented and fitted based on the adjusted ground segmentation height threshold.

[0036] As an optional implementation, the method further includes:

[0037] Real-time obstacle avoidance path planning is performed based on the targets in the dynamic obstacle layer.

[0038] The static obstacle layer is used to provide boundary constraint information for the path planning.

[0039] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 is a flowchart of an environmental perception method based on lidar and high-precision map provided in an embodiment of this application;

[0042] Figure 2 is a flowchart of a method for dividing a region of interest according to an embodiment of this application;

[0043] Figure 3 is a flowchart of a method for dividing a region of non-interest provided in an embodiment of this application;

[0044] Figure 4 is a flowchart of a target recognition and classification method provided in an embodiment of this application;

[0045] Figure 5 is a flowchart of a target partitioning method provided in an embodiment of this application;

[0046] Figure 6 is a flowchart of a method for optimizing the classification and tracking results of targets according to an embodiment of this application;

[0047] Figure 7 is a flowchart of a slope road surface identification and optimization method provided in an embodiment of this application;

[0048] Figure 8 is a flowchart of an environmental perception application method provided in an embodiment of this application. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] The following will describe in detail an environmental perception method based on LiDAR and high-precision maps provided in this application embodiment, with reference to specific implementation methods. Figure 1 is a flowchart of an environmental perception method based on LiDAR and high-precision maps provided in this application embodiment. As shown in Figure 1, the specific steps are as follows:

[0051] Step 101: Collect environmental point cloud data in real time using lidar.

[0052] In practice, LiDAR (Light Detection and Ranging) acquires three-dimensional information about objects in the environment by emitting laser pulses and receiving reflected signals, generating point cloud data. Point cloud data contains information such as the position and shape of target objects and is the foundation of environmental perception. For example, in autonomous driving, a 32-line LiDAR installed on the roof or body of the vehicle collects point cloud data within a 360° range around the vehicle at a scanning frequency of 10Hz.

[0053] Step 102: Divide the region of interest based on the high-precision map.

[0054] In implementation, road boundary information provided by high-precision maps is used to divide the area around the vehicle into regions of interest (ROI) (e.g., within the road) and non-ROI (e.g., outside the road). ROIs can be processed more efficiently by focusing resources on them, reducing the computational burden on unrelated areas. For example, global road boundary data from a high-precision map can be read and combined with the vehicle's positioning information to extract a local road area 50 meters in front of the vehicle and 5 meters to its left and right as the ROI. Based on the spatial distribution of point cloud data, points within the ROI are further divided into road points and non-road points.

[0055] As an optional implementation, Figure 2 is a flowchart of a method for dividing a region of interest according to an embodiment of this application. As shown in Figure 2, the specific steps for dividing the region of interest based on a high-precision map in step 102 are as follows:

[0056] Step 201: Read the global road boundary information from the high-precision map.

[0057] In implementation, high-precision maps contain geometric information about road boundaries (such as lane lines and curbs), which is stored as vector data. By parsing high-precision maps, road boundaries can be extracted globally. For example, after a vehicle starts, the system connects to the high-precision map database and reads the global road boundary information for the current area, including the shape, length, width, and adjacency relationships of each road. Furthermore, the system can extract the latitude and longitude coordinates of road boundaries and road types (such as main roads, auxiliary roads, and sidewalks) from OSM maps.

[0058] Step 202: Based on the vehicle's location, obtain local road boundary information within a preset range near the vehicle.

[0059] In implementation, by combining the vehicle's real-time location information (such as GPS coordinates or IMU data), a local area related to the vehicle's position is filtered out from the global road boundary. Limiting the processing range can significantly reduce the computational load. For example, when the vehicle's real-time location shows that it is in the middle of a road segment, the road boundary information within a range of 50 meters in front and behind the vehicle and 5 meters to the left and right is extracted with the vehicle as the center to generate the spatial range of the local road.

[0060] Step 203: Construct the point cloud data boundary corresponding to the local road boundary.

[0061] In implementation, point cloud data collected by LiDAR can be matched with local boundary information in a high-precision map to generate point cloud data boundaries corresponding to road boundaries, which are used to accurately delineate regions of interest. For example, all lanes within a preset range near the vehicle can be acquired. First, a rough judgment can be made as to whether the point cloud falls within the extreme range of a certain lane. Then, a more detailed judgment can be made as to whether the point cloud is within the polygon boundary point cloud frame, thus distinguishing the point clouds inside and outside the road. Alternatively, within a local area, the coordinates of the point cloud collected by LiDAR can be projected onto the road boundary coordinates of the high-precision map to generate a corresponding boundary point cloud. A projection algorithm can be used to align the LiDAR point cloud with the main lane boundary and mark points outside the boundary as invalid points.

[0062] Step 204: Define the area within the local road boundary as the region of interest.

[0063] In implementation, the area within the local road boundary is typically the main range of vehicle travel and is a key area for environmental perception. By defining this area as the region of interest (ROI), the accuracy and efficiency of subsequent point cloud data processing and obstacle detection can be optimized. For example, all point cloud data within the local boundary can be designated as the ROI and labeled, such as "lane-in-lane point cloud." In subsequent processing, only these points are clustered and detected, such as identifying dynamic targets like vehicles and pedestrians, while point clouds outside the boundary are ignored.

[0064] As an optional implementation, Figure 3 is a flowchart of a method for dividing a region of non-interest according to an embodiment of this application. As shown in Figure 3, the specific steps are as follows:

[0065] Step 301: Define the area outside the local road boundary as the region of non-interest.

[0066] In implementation, areas outside the local road boundaries are typically areas inaccessible to vehicles or areas of no concern, such as sidewalks and buildings. Defining regions of interest (ROIs) reduces the scope and computational burden of point cloud data processing, improving efficiency and stability. For example, the system can filter point clouds collected by LiDAR based on the local boundary information of the vehicle's current road. Point clouds extending beyond the road boundaries (such as trees, pedestrians, or fixed buildings spanning the road) are marked as ROIs. For instance, when a vehicle is on a two-lane straight road, the system will directly mark the point clouds of the green belt outside the road boundary as "ROIs."

[0067] Step 302: Assign targets within the non-interest area to the static obstacle layer.

[0068] In practice, targets within non-interest areas (NIIs) typically have little impact on vehicle decisions (such as pedestrians or stationary objects outside the road). Classifying these targets as static obstacles helps prevent them from occupying drivable areas or falsely triggering obstacle avoidance mechanisms. For example, the system can detect targets within NIIs using point cloud clustering algorithms. For instance, for trees located at the road edge, the system treats their complete point cloud as a static obstacle, labeling its boundary position and size, but without dynamic tracking. Similarly, building point clouds within NIIs (such as walls, billboards, etc.) are also directly classified as static obstacles, storing their 3D geometric information to provide scene boundary references.

[0069] Step 103: Perform target identification and classification on the environmental point cloud data within the region of interest.

[0070] In implementation, the geometric features of point clouds (such as height and density) and the motion characteristics of targets (such as speed and trajectory) are used to classify objects within the region of interest, initially distinguishing targets such as vehicles and pedestrians, providing a basis for further obstacle processing. For example, point cloud data can be grouped using clustering algorithms, and the size, density, and height of each group can be calculated. If the point cloud height is within the range of 0.5-2.5 meters, the density is high, and the speed is close to 0, it is identified as a stationary vehicle. If the height is less than 2 meters and the speed is low, it is identified as a pedestrian target.

[0071] As an optional implementation, Figure 4 is a flowchart of a target recognition and classification method provided in an embodiment of this application. As shown in Figure 4, the specific steps of target recognition and classification of environmental point cloud data in the region of interest in step 103 are as follows:

[0072] Step 401: Initially identify the target type based on the target's motion characteristics and geometric features.

[0073] In practice, the motion characteristics (such as speed and direction) and geometric features (such as shape and size) of the target are the main criteria for distinguishing pedestrians, vehicles, and stationary objects. By analyzing the motion trajectory and morphological distribution of point clouds, the target type can be preliminarily identified, providing a foundation for subsequent optimization. The system can cluster and perform motion analysis on point clouds within a region of interest, and can calculate the speed and direction of motion of point cloud clusters using time-series data. For example, if a point cloud cluster has a continuous moving trajectory and a speed between 5-10 km / h, it is preliminarily identified as a pedestrian; if the speed is higher than 30 km / h, it may be a vehicle. The system can also determine the target type based on the three-dimensional dimensions and density of the point cloud. For example, a point cloud cluster with a length greater than 4 meters and a width of about 2 meters may be a small vehicle. A point cloud cluster with a height less than 2 meters may be a pedestrian.

[0074] Step 402: Verify and optimize the preliminary identification results based on the regional attributes of the high-precision map.

[0075] In implementation, high-precision maps provide regional attributes (such as lanes, sidewalks, intersections, etc.), which can be used to verify point cloud target recognition results. For example, pedestrians are usually located on sidewalks, while vehicles are mostly in lanes. Combining regional attributes can correct misclassifications and improve recognition accuracy. For example, the system can perform secondary judgments on the recognition results based on the regional attributes of the high-precision map. For instance, if a pedestrian target initially identified is located in a lane, it may be a misclassification, and its shape and speed can be reassessed to correct the result. If a target is located on a ramp, the system can optimize its trajectory by combining the ramp's inclination information to ensure classification stability. In intersection areas, the system further subdivides multiple targets within the area and determines whether they are vehicle or pedestrian queues through trajectory analysis.

[0076] Step 104: Based on the target's motion attributes, geometric characteristics, and point cloud density, classify the target within the region of interest into a dynamic obstacle layer or a static obstacle layer.

[0077] In practice, by analyzing the motion state and geometric characteristics of targets, dynamic targets (such as moving vehicles) are classified into a dynamic obstacle layer, while stationary targets or objects of no interest (such as fences or trees) are classified into a static obstacle layer. For example, if the motion attributes of a detected target show a speed greater than 1 m / s and its direction is consistent with the road, it is classified into the dynamic obstacle layer. If the target size is less than 0.3 meters or the point cloud density is low, it is directly classified into the static obstacle layer. For example, if a tree trunk crossing the road is detected with a height greater than 0.8 meters but a density below a threshold, it is classified into the static obstacle layer.

[0078] As an optional implementation, Figure 5 is a flowchart of a target segmentation method provided in an embodiment of this application. As shown in Figure 5, the specific steps in step 104 of segmenting the target within the region of interest into a dynamic obstacle layer or a static obstacle layer based on the target's motion attributes, geometric characteristics, and point cloud density are as follows:

[0079] Step 501: Assign targets with dynamic motion attributes to the dynamic obstacle layer.

[0080] In implementation, dynamic targets are obstacles that significantly affect vehicle movement, such as pedestrians and vehicles. By analyzing the motion of point clouds across consecutive frames, their motion attributes can be identified, and moving targets can be categorized into a dynamic obstacle layer for subsequent tracking and obstacle avoidance planning. For example, time-series point cloud data from LiDAR can be used to analyze the continuous positional changes of point cloud clusters. If a point cloud cluster exhibits a significant velocity (e.g., ≥1 m / s), it is identified as a dynamic target, possibly a moving pedestrian or vehicle. Targets with dynamic motion characteristics (such as detected pedestrians, vehicles, or bicycles) are stored in the dynamic obstacle layer.

[0081] Step 502: Assign targets with static motion attributes to the static obstacle layer.

[0082] In implementation, static targets such as buildings, streetlights, or parked vehicles do not cause real-time interference to dynamic obstacle avoidance. Classifying point cloud targets with static motion attributes into a static obstacle layer can effectively reduce the computational burden of dynamic obstacle avoidance. For example, the system can perform motion trajectory analysis on point cloud clusters. If a point cloud target has not moved within multiple frames and its velocity is less than 0.1 m / s, it is judged as a static target, such as a parked vehicle or a fixed road obstacle. The system can then classify such targets into a static obstacle layer and record their positions to assist in path boundary planning.

[0083] Step 503: Small targets with point cloud density below a preset threshold are classified into the static obstacle layer.

[0084] In practice, small targets with low point cloud density are often misidentified or irrelevant distractions (such as leaves, pebbles, etc.). Classifying them into the static obstacle layer can avoid affecting the accuracy of path planning. For example, the system can calculate the point cloud density and size for each point cloud cluster. If the point cloud density of a target is lower than a set threshold (e.g., 10 points / m³) and the size is less than 0.5m³, the system determines it to be a non-threatening object (such as a small animal or garbage). This target is classified into the static obstacle layer and marked as a low-priority obstacle.

[0085] Step 504: Assign the abnormal point cloud targets that failed to be detected to the static obstacle layer.

[0086] In implementation, anomalous point cloud targets may fail to be classified due to data loss, noise, or irregular shapes. Classifying these targets into a static obstacle layer can prevent misclassification from interfering with dynamic obstacle avoidance. For example, for point clouds that fail to cluster or fail to detect, if a target point cloud lacks a clear shape or trajectory, the system can classify it as an anomalous target (such as complex tree canopies or low-quality point clouds). The system can classify it into a static obstacle layer, retaining only its original shape and position to aid in environmental modeling. For instance, a point cloud of tree branches crossing a road can be classified into a static obstacle layer to avoid erroneously triggering obstacle avoidance mechanisms.

[0087] As an alternative implementation, the original point cloud shape of the target is preserved in the static obstacle layer.

[0088] In implementation, preserving the original point cloud shape of static obstacles helps to accurately represent the spatial characteristics of targets, thereby improving the accuracy of environmental modeling. This is especially true in scenarios with irregularly shaped obstacles (such as tree canopies, billboards, or irregularly shaped roadblocks), as it avoids simplified polygonal envelopes occupying too much drivable area, optimizing path planning and decision-making. For example, the system can label point cloud clusters in the static obstacle layer, distinguishing them into targets such as buildings, trees, and parked vehicles. For instance, if a roadside tree canopy is detected with an irregularly distributed point cloud that cannot be accurately fitted using a geometric model, the system avoids envelope approximation of static targets and stores only the complete point cloud data in the static obstacle layer. For the aforementioned tree canopy, all its point cloud coordinates are preserved, and the height and coverage of the lowest point are recorded. During path planning, dynamic obstacles are updated in real time, while the point cloud data from the static obstacle layer can be directly used to define the boundaries of the obstacle avoidance area. When a vehicle approaches a tree canopy, the system can determine whether the canopy's lowest point affects the vehicle's passage based on the canopy's height. If it is below the vehicle's height limit, it is marked as an area to be avoided; otherwise, it is only used as background data.

[0089] Step 105: Optimize the classification and tracking results of the target based on the road attribute information of the high-precision map, and assist in path planning and decision-making.

[0090] In implementation, road attributes (such as ramps, straight sections, and sidewalks) in high-precision maps provide prior information for target classification and are used to correct tracking results and optimize vehicle path planning and obstacle avoidance strategies. For example, straight section information can be used to filter out target trajectory jitter, and ramp information can improve the accuracy of ground segmentation. For example, the speed of pedestrian targets can be limited by combining sidewalk attributes; if a target's speed is detected to be too high and its trajectory deviates from the sidewalk, it is corrected to a non-pedestrian target. If ramp information shows that the current road is inclined at 10 degrees, the height threshold for ground segmentation is dynamically adjusted to 0.3 meters to avoid misclassifying ramps as obstacles. At the same time, targets in the static obstacle layer, such as roadside railings, are avoided when planning paths.

[0091] As an optional implementation, Figure 6 is a flowchart of a method for optimizing the classification and tracking results of targets according to an embodiment of this application. As shown in Figure 6, the road attribute information of the high-precision map includes slope attributes, straight road attributes, and sidewalk attributes. The specific steps of optimizing the classification and tracking results of targets based on the road attribute information of the high-precision map in step 105 are as follows:

[0092] Step 601: Based on the sidewalk attributes, verify and limit the speed of pedestrian targets.

[0093] In implementation, pedestrian walkway attributes provide prior constraints for target classification. By defining the pedestrian walkway's range in a high-precision map, the speed of pedestrian targets can be limited, preventing the misidentification of slow-moving objects (such as carts or stationary targets) as dynamic pedestrians and reducing erroneous obstacle avoidance or path planning triggers. For example, if the system detects a moving target on the pedestrian walkway at 8 km / h, based on the pedestrian walkway's speed limit (e.g., 10 km / h), the target is identified as a pedestrian. If the speed exceeds the normal pedestrian walkway speed limit (e.g., 30 km / h), it is reclassified as a non-pedestrian target, possibly a cyclist or vehicle. In path planning, interactions with pedestrian targets are restricted; for example, when a vehicle approaches a pedestrian walkway, a speed condition is set to prioritize pedestrian avoidance. If multiple pedestrian targets are detected on the pedestrian walkway, the system adjusts the planned path, either slowing down to pass or choosing a non-interfering area to detour.

[0094] Step 602: Based on the ramp attributes, identify and optimize the ramp surface.

[0095] In implementation, the unique terrain of slope areas can easily lead to ground missegmentation and target misclassification. For example, the height deviation of the LiDAR point cloud may occur due to the slope inclination, causing the ground to be misidentified as an obstacle. Optimizing ground recognition through slope attributes improves segmentation accuracy. For instance, the system can adjust the ground segmentation height threshold of the LiDAR point cloud based on slope inclination information (e.g., a 10° slope) in a high-precision map, including slope point clouds in ground classification. For example, the ground threshold for ordinary road sections is 0.2m, while it is increased to 0.5m on slopes to adapt to slope variations. When a vehicle enters a slope, the slope surface model is refitted, filtering out non-ground targets (such as parked vehicles or roadside guardrails).

[0096] Step 603: Based on the straight-line attribute, verify the target direction information and optimize the tracking trajectory.

[0097] In implementation, the straight-line attribute provides directional constraints for target tracking, reducing trajectory jitter caused by sensor errors or environmental interference. For example, in a straight-line scenario, the target's movement direction should be consistent with the road direction; directional changes outside this range can be considered abnormal trajectories and corrected. For instance, on a straight-line, a vehicle is detected moving at 30°, while the straight-line direction is 0°. The system determines its trajectory is abnormal and corrects it to the road direction. If the target's direction deviates too much (e.g., 90°), it is marked as a misclassified or irrelevant target. Directional constraints are applied to the target's historical trajectory; for example, a vehicle's trajectory on a straight-line should be a smooth straight line. If the trajectory exhibits unreasonable left-right swaying, it is adjusted to a smooth path using a filtering algorithm. When the vehicle is tracking the vehicle ahead on a straight-line, the optimized target trajectory improves the vehicle's path stability, avoiding unnecessary braking or acceleration caused by target trajectory jitter.

[0098] As an optional implementation, Figure 7 is a flowchart of a slope road surface identification and optimization method provided in an embodiment of this application. As shown in Figure 7, the specific steps of identifying and optimizing the slope road surface based on slope attributes in step 602 are as follows:

[0099] Step 701: Obtain the inclination angle of the ramp.

[0100] In implementation, the slope attributes include the slope's inclination angle, serving as a core parameter describing terrain features. By reading the inclination angle, the system can adjust the point cloud data processing algorithm to adapt to the specific terrain of slope sections, avoiding misjudgments in ground segmentation. For example, the system can read the road attributes of the vehicle's current location through a high-precision map interface to obtain the inclination angle of the slope area, such as a 10° uphill slope. If the map contains the start and end coordinates of the slope, it can also combine the vehicle's positioning to further confirm the specific slope segment. When the vehicle enters a new road segment, the slope inclination angle is updated in real time. For example, if the current slope is a 10° uphill slope, and after driving for a while, it enters a 15° steep slope, the system dynamically adjusts the relevant parameters.

[0101] Step 702: Dynamically set the ground segmentation height threshold according to the tilt angle.

[0102] In implementation, slope inclination can cause a shift in the height of LiDAR point cloud data. By adjusting the ground segmentation height threshold, ground and non-ground targets (such as vehicles and guardrails) can be accurately distinguished, avoiding missegmentation or omission of ground features. The system can dynamically adjust the threshold based on the slope inclination angle. For example, when the slope inclination angle is 0°, the ground height threshold is 0.2m; when the slope inclination angle is 10°, the threshold increases to 0.4m; and when the inclination angle reaches 20°, the threshold is set to 0.6m. When the system detects a slope inclination angle of 12°, it dynamically sets the ground height threshold to 0.45m. Points in the LiDAR point cloud with a height below 0.45m are identified as ground, preventing the outer edge of the slope from being mistakenly marked as a ground area.

[0103] Step 703: Perform segmented fitting of the sloping road surface based on the adjusted ground segmentation height threshold.

[0104] In implementation, a single planar model cannot accurately describe the complex terrain of sloping roads. By segmenting and fitting the ground, the system can generate a more accurate sloping road surface model, improving perception accuracy and providing reliable support for subsequent path planning. The system can divide the sloping road surface into several small segments, fitting a plane within each segment based on local point cloud data. For example, the slope can be divided into uniform 5-meter segments, with each segment fitted with a slope. By stitching together these segments, a complete sloping road surface model is generated. If, during the fitting process, a segment's point cloud deviates from the overall trend (such as abrupt steps), the system can automatically adjust the fitting parameters for that segment to smoothly connect it with adjacent segments. The system can also verify the fitting results to ensure they match the slope attributes of the high-precision map, such as consistent start and end points and inclination angles.

[0105] As an optional implementation, Figure 8 is a flowchart of an environmental perception application method provided by an embodiment of this application. As shown in Figure 8, the application of dynamic obstacle layers and static obstacle layers is as follows:

[0106] Step 801: Perform real-time obstacle avoidance path planning based on the target in the dynamic obstacle layer.

[0107] In implementation, targets in the dynamic obstacle layer are moving objects, such as vehicles or pedestrians, whose trajectories are uncertain. By tracking these targets in real time, the system can predict their movement trends and dynamically adjust obstacle avoidance paths to ensure driving safety and smoothness. Specifically, the system predicts the trajectories of targets in the dynamic obstacle layer. For example, a pedestrian is moving along the sidewalk at a speed of 1.5 m / s, and a vehicle ahead is traveling in a straight line at a speed of 10 m / s, with the possibility of changing lanes. Based on the target's movement trend, the system adds a dynamic obstacle avoidance zone to the vehicle's path planning to avoid collisions. The system can also generate obstacle avoidance paths based on the real-time position and trajectory of dynamic obstacles. For example, if there is a slow-moving vehicle 10 meters ahead, the system plans a path to change lanes to the left. If the path cannot meet the safe distance requirements, the system reduces the vehicle's speed to avoid a collision. When the trajectory of a dynamic obstacle changes (such as sudden stopping or turning), the system updates the obstacle avoidance path in real time. For example, if a pedestrian suddenly accelerates across the road on a zebra crossing, the vehicle will immediately adjust its deceleration or stopping strategy.

[0108] Step 802: Use the static obstacle layer to provide boundary constraint information for path planning.

[0109] In implementation, the targets in the static obstacle layer are stationary objects (such as guardrails, buildings, or road signs) whose positions are relatively stable. By providing boundary information of static obstacles for path planning, the system can ensure that path planning is carried out within a reasonable range, avoiding collisions or deviation from the road. Specifically, the system can extract the position and outline data of obstacles from the static obstacle layer. For example, the guardrail on the right side of the road is 1.5 meters away from the vehicle, the width of the stationary obstacle is 0.8 meters, and it is located 1 meter to the left of the center of the lane. In path planning, the boundaries of static obstacles are set as impassable areas. For example, the left boundary of the planned path is the lane line, and the right boundary is the outer edge of the guardrail. This ensures that the vehicle always travels between the two boundaries and does not deviate from the safe area. Furthermore, when the vehicle is driving on a narrow road, the system can use the static obstacle layer to ensure that the planned path stays close to the boundaries while avoiding collisions. When the system detects a construction fence ahead, it automatically adjusts the planned path, allowing the vehicle to detour and return to the safe driving area.

[0110] This application provides an environmental perception method based on LiDAR and high-precision maps. The method includes: real-time acquisition of environmental point cloud data using LiDAR; division of regions of interest (ROIs) based on the high-precision map; target identification and classification within the ROIs of the environmental point cloud data; classification of targets within the ROIs to either a dynamic obstacle layer or a static obstacle layer based on their motion attributes, geometric characteristics, and point cloud density; and optimization of target classification and tracking results based on road attribute information from the high-precision map, which also assists in path planning and decision-making. This application, by combining LiDAR point cloud data and high-precision map information, achieves accurate environmental perception, reduces noise interference in traditional point cloud processing, and maintains high stability, especially in complex scenarios (such as ramps and intersections). By dividing the ROIs, clustering and detection are performed only on point clouds within road boundaries, avoiding redundant calculations on data in non-critical areas and improving computational efficiency. Based on the target's motion attributes, geometric characteristics, and point cloud density, obstacle types are dynamically distinguished, minimizing the drivable area occupied by static obstacles while improving the accuracy of dynamic obstacle detection, providing more reliable information for vehicle decision-making. By leveraging road attribute information (such as ramps, straight sections, and sidewalks) from high-precision maps, target tracking and ground segmentation algorithms are effectively optimized, improving the system's applicability and accuracy in special scenarios (such as ramp fitting or straight section target direction correction). A dynamic obstacle layer provides real-time dynamic information for path planning, while a static obstacle layer defines boundary constraints by preserving point cloud shapes, avoiding misjudgments of driving areas and improving the feasibility and safety of obstacle avoidance paths. Combining target shape, speed, and other characteristics with high-precision map attributes, the target classification algorithm is optimized, especially in the absence of visual assistance, effectively distinguishing vehicles from pedestrians and enhancing the intelligence level of environmental perception. In scenarios such as tree canopies obstructing roads, irregularly shaped objects crossing boundaries, or physically isolated targets (such as elevated multi-level roads), the static obstacle layer's segmentation strategy effectively avoids misjudgments and unnecessary obstacle avoidance actions, improving vehicle stability and safety. The method in this application's embodiments is not only applicable to autonomous driving scenarios but can also be extended to unmanned delivery vehicles, security monitoring, and other fields, providing an effective solution for environmental perception in multiple scenarios.

[0111] It should be understood that although the steps in the flowcharts of Figures 1 to 8 are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in Figures 1 to 8 may include multiple steps or multiple stages, which are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0112] It is understood that the same / similar parts between the various embodiments of the methods described above in this specification can be referred to each other. Each embodiment focuses on the differences from other embodiments, and relevant parts can be referred to the description of other method embodiments.

[0113] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0114] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0115] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0116] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. An environmental perception method based on lidar and high-precision maps, characterized in that, The method includes: Real-time acquisition of environmental point cloud data via lidar; Regions of interest are defined based on high-precision maps; Target identification and classification are performed on the environmental point cloud data within the region of interest; Based on the target's motion attributes, geometric characteristics, and point cloud density, the target within the region of interest is divided into a dynamic obstacle layer or a static obstacle layer. The classification and tracking results of targets are optimized based on the road attribute information of the high-precision map, and path planning and decision-making are assisted.

2. The method according to claim 1, characterized in that, The division of regions of interest based on high-precision maps includes: Read the global road boundary information from the high-precision map; Based on the vehicle's location, obtain local road boundary information within a preset range near the vehicle; Construct the point cloud data boundary corresponding to the local road boundary; The region within the local road boundary is defined as the region of interest.

3. The method according to claim 2, characterized in that, The method further includes: The area outside the local road boundary is defined as the region of non-interest. The targets within the non-interest area are assigned to the static obstacle layer.

4. The method according to claim 1, characterized in that, The step of identifying and classifying targets in the environmental point cloud data within the region of interest includes: Preliminary target type identification based on target motion characteristics and geometric features; The preliminary identification results are verified and optimized based on the regional attributes of the high-precision map.

5. The method according to claim 1, characterized in that, The step of classifying targets within the region of interest into a dynamic obstacle layer or a static obstacle layer based on their motion attributes, geometric characteristics, and point cloud density includes: The targets with dynamic motion attributes are assigned to the dynamic obstacle layer; The target with static motion attributes is assigned to the static obstacle layer; Small-sized targets with point cloud density below a preset threshold are assigned to the static obstacle layer; The abnormal point cloud targets that failed to be detected are assigned to the static obstacle layer.

6. The method according to claim 5, characterized in that, The method further includes: The original point cloud shape of the target is preserved in the static obstacle layer.

7. The method according to claim 1, characterized in that, The road attribute information of the high-precision map includes slope attributes, straight road attributes, and sidewalk attributes.

8. The method according to claim 7, characterized in that, The classification and tracking results of the target optimization based on the road attribute information of the high-precision map include: Based on the pedestrian walkway attributes, the speed of pedestrian targets is verified and limited; Based on the aforementioned ramp attributes, the ramp surface is identified and optimized; Based on the straight-line attribute, the target direction information is verified and the tracking trajectory is optimized.

9. The method according to claim 8, characterized in that, The process of identifying and optimizing the slope surface based on the slope attributes includes: Obtain the slope angle of the ramp; The ground segmentation height threshold is dynamically set according to the tilt angle; The slope road surface is segmented and fitted based on the adjusted ground segmentation height threshold.

10. The method according to claim 1, characterized in that, The method further includes: Real-time obstacle avoidance path planning is performed based on the targets in the dynamic obstacle layer. The static obstacle layer is used to provide boundary constraint information for the path planning.

Images