Perception method and system for target detection

Abstract

The embodiment of the invention provides a sensing method and system for target detection. The method comprises the following steps: acquiring 3D point cloud data in a predetermined detection range acquired by an acquisition device; inputting the 3D point cloud data into a 3D point cloud target detection algorithm unit, and determining an output result of a 3D detection bounding box of a detection target by the 3D point cloud target detection algorithm unit; inputting the 3D point cloud data into a 3D point cloud segmentation clustering algorithm unit, and determining an output result of a 3D clustering bounding box of a detection target by the 3D point cloud segmentation clustering algorithm unit; and fusing an output result of the 3D point cloud target detection algorithm unit and an output result of the 3D point cloud segmentation clustering algorithm unit, and determining an output result of fusion information. Through the processing scheme disclosed by the invention, the sensing accuracy of the solid-state laser radar is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a perception method and system for target detection. Background Art

[0002] In the field of high-level autonomous driving, lidar is considered to be one of the key sensors. Existing lidars are mainly divided into two categories: mechanical lidar and solid-state lidar.

[0003] Mechanical LiDAR achieves 360° fixed scanning through internal rotating parts. However, due to the moving parts inside the mechanical LiDAR, the radar has poor reliability, short life, and high cost. As a result, mechanical LiDAR is currently unable to pass automotive-grade certification and is commonly used in non-mass-produced high-end autonomous driving vehicles for function and algorithm verification.

[0004] Solid-state LiDAR does not require rotating and moving scanning parts, which gives it certain advantages in stability and reliability, and it is increasingly being adapted to mass-produced models. Solid-state LiDAR uses an optical phased array and Flash integrated design, and mainly relies on the reflection and reception of light waves to achieve detection purposes. Solid-state LiDAR naturally has certain advantages and disadvantages. Its advantages are simple structure without rotating parts, high scanning accuracy, good controllability, and fast scanning speed. But at the same time, solid-state LiDAR will also have problems such as limited scanning angles, sidelobe effects, and high processing difficulty in a short period of time. Despite this, for mass-produced automotive LiDARs, solid-state radars will still become the future installation trend due to their stability, reliability, and low cost.

[0005] In addition, with the technical iteration and update of the academic community in the field of 3D point cloud perception, there are many different algorithm processing ideas in the field of 3D point cloud perception: compressing the point cloud into a 2D image to perceive the target, but this will lose the height information characteristics of the point cloud to a certain extent; converting the point cloud into columnar voxels, extracting features and compressing them into 2D images, and using the image feature extraction network to detect the 3D frame of the target; processing the point cloud into gridded voxels for 3D feature extraction and target perception, with a certain balance between processing speed and accuracy; directly performing point-level sampling, feature extraction and aggregation based on the original point cloud, and then implementing CXWL2022015A-A57

[0006] However, this processing method is sensitive to the amount of point cloud data. The larger the amount of point cloud data, the longer the processing time.

[0007] Different lidar target detection methods have different advantages and disadvantages. In actual use, the industry's perception solutions are different based on the characteristics of different lidars. At present, the most widely used perception application based on solid-state lidar is the single-frame point cloud detection solution based on deep learning algorithm, which converts the acquired single-frame point cloud into columnar voxels, extracts features and compresses them into 2D images, and uses the image feature extraction network to perform 3D frame detection of the target.

[0008] However, it should be noted that the existing publicly available solid-state LiDAR point cloud perception solutions only detect targets based on single-frame point cloud data. On the one hand, they ignore the non-repetitive scanning characteristics of solid-state LiDAR. On the other hand, only using the target detection algorithm has limited perception performance for target obstacles and cannot achieve more accurate environmental perception. The algorithm for target detection based on single-frame point cloud data obviously cannot give full play to the inherent perception advantages of solid-state LiDAR to a large extent.

[0009] Therefore, there is a need for a target detection perception method and system that can improve the performance and detection accuracy of existing lidar. Summary of the invention

[0010] In view of this, an embodiment of the present disclosure provides a perception method for target detection, which at least partially solves the problems existing in the prior art.

[0011] In a first aspect, an embodiment of the present disclosure provides a perception method for target detection, characterized in that the method includes:

[0012] Acquire 3D point cloud data within a predetermined detection range collected by a collection device;

[0013] Inputting the 3D point cloud data into a 3D point cloud target detection algorithm unit, and the 3D point cloud target detection algorithm unit determines an output result of a 3D detection bounding box of the detection target;

[0014] Inputting the 3D point cloud data into a 3D point cloud segmentation and clustering algorithm unit, and allowing the 3D point cloud segmentation and clustering algorithm unit to determine an output result of a 3D clustering bounding box of a detection target; and

[0015] The output result of the 3D point cloud target detection algorithm unit and the output result of the 3D point cloud segmentation and clustering algorithm unit are fused, and the output result of the fusion information is determined.

[0016] CXWL2022015A-A57

[0017] According to a specific implementation method of an embodiment of the present disclosure, when the 3D point cloud target detection algorithm unit decides to output a 3D detection bounding box of the detection target, and the 3D point cloud segmentation and clustering algorithm unit decides to output a 3D clustering bounding box of the detection target, the 3D overlap of the 3D detection bounding box and the 3D clustering bounding box is calculated, and based on a comparison between the 3D overlap and a preset overlap threshold, the output result of the fusion information is determined.

[0018] According to a specific implementation manner of an embodiment of the present disclosure, when the 3D overlap is greater than a preset overlap threshold, the output result of the fusion information is the 3D detection bounding box; and when the 3D overlap is less than the preset overlap threshold, the Euclidean distance between the 3D detection bounding box and the 3D clustering bounding box is calculated, wherein, when the Euclidean distance is less than a preset Euclidean distance threshold and the confidence of the 3D detection bounding box is greater than a preset confidence threshold, the output result of the fusion information is the 3D detection bounding box, and when the Euclidean distance is greater than the preset Euclidean distance threshold, the output result of the fusion information is the 3D detection bounding box.

[0019] According to a specific implementation of the embodiment of the present disclosure, the method further includes:

[0020] When the 3D point cloud target detection algorithm unit decides not to output the 3D detection bounding box of the detection target, and the 3D point cloud segmentation clustering algorithm unit decides to output the 3D clustering bounding box of the detection target, the number of points of the detection target of the 3D clustering bounding box is judged, wherein,

[0021] When the number of points is greater than a preset point threshold, determining that the detected target is an obstacle; and

[0022] When the number of points is less than a preset point threshold, the detection target is determined to be a noise point or a misidentified obstacle.

[0023] According to a specific implementation of the embodiment of the present disclosure, the method further includes:

[0024] The detection target is classified according to a preset classification algorithm, and the classification includes classified obstacles and unclassified obstacles.

[0025] In a second aspect, an embodiment of the present disclosure provides a perception system for target detection, the system comprising:

[0026] An acquisition module is configured to acquire 3D point cloud data within a predetermined detection range acquired by an acquisition device;

[0027] The detection module is configured to input the 3D point cloud data into the 3D point cloud target detection algorithm unit CXWL2022015A-A57

[0028] The 3D point cloud target detection algorithm unit determines the output result of the 3D detection bounding box of the detection target;

[0029] A clustering module is configured to input the 3D point cloud data into a 3D point cloud segmentation clustering algorithm unit, and the 3D point cloud segmentation clustering algorithm unit determines the output result of the 3D clustering bounding box of the detection target; and a fusion module is configured to fuse the output result of the 3D point cloud target detection algorithm unit and the output result of the 3D point cloud segmentation clustering algorithm unit, and determine the output result of the fusion information.

[0030] According to a specific implementation of the embodiment of the present disclosure, the system further includes: a classification module, configured to classify the detection target according to a preset classification algorithm, and the classification includes classified obstacles and unclassified obstacles.

[0031] In a third aspect, an embodiment of the present disclosure further provides an electronic device, the electronic device comprising:

[0032] at least one processor; and,

[0033] a memory communicatively connected to the at least one processor; wherein,

[0034] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the perception method for target detection in the aforementioned first aspect or any implementation of the first aspect.

[0035] In a fourth aspect, an embodiment of the present disclosure further provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the perception method for target detection in the aforementioned first aspect or any implementation of the first aspect.

[0036] In the fifth aspect, the embodiments of the present disclosure also provide a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the perception method for target detection in the aforementioned first aspect or any implementation of the first aspect.

[0037] The perception method for target detection in the disclosed embodiment combines multi-frame point cloud data for target detection, segmentation, clustering and target fusion, and extracts features by combining multi-frame point cloud information, which is more conducive to the information representation of point cloud. The 3D point cloud target detection algorithm is used to detect and identify obstacles that are regular and easy to classify. The 3D point cloud target segmentation and clustering algorithm is used to extract all obstacles on the road surface, and CXWL2022015A-A57

[0038] All obstacles are unified and integrated to finally obtain the real obstacle 3D information of the scene. The solution of the embodiment of the present invention can significantly improve the perception performance and detection accuracy of the solid-state laser radar. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A flowchart of a perception method for target detection provided by an embodiment of the present disclosure;

[0041] Figure 2 A schematic diagram of a method for performing target detection at time t based on a solid-state laser radar provided in an embodiment of the present disclosure;

[0042] Figure 3 A schematic diagram of a perception system for target detection provided by an embodiment of the present disclosure; and

[0043] Figure 4 A schematic diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0044] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0045] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0046] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0047] CXWL2022015A-A57

[0048] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0049] In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the aspects described may be practiced without these specific details. At present, there is no comprehensive 3D point cloud perception solution for solid-state laser radar. Therefore, the present invention proposes a perception method for target detection, which can significantly improve the perception performance of solid-state laser radar by performing feature extraction, target detection and fusion on point clouds of multiple consecutive frames, and performing voxel feature extraction, bounding box prediction, semantic feature extraction, ground point cloud filtering and target clustering on the point clouds of the current frame and previous frames.

[0050] Figure 1A flowchart of a perception method 100 for target detection provided in an embodiment of the present disclosure.

[0051] like Figure 1 As shown, at step S110, 3D point cloud data within a predetermined detection range collected by a collection device is acquired.

[0052] For fixed LiDAR scenes, the distance that the target moves in the scene (or the predetermined detection range) in a short period of time is limited, and the distribution of the target position is relatively close, but the scene data scanned by different frames are not repeated. Therefore, from the data perspective, each frame of point cloud data has different target scanning characteristics. For vehicle-mounted non-fixed LiDAR scenes, the multi-frame point cloud data can be converted to the same radar coordinate system by introducing the vehicle's own motion parameters to keep it consistent with the fixed LiDAR scene.

[0053] More specifically, multiple frames of 3D point cloud data in a predetermined time sequence within a predetermined detection range are collected.

[0054] The embodiment of the present invention selects to collect three consecutive frames of point cloud data in a time series, but this does not constitute a limitation to the present invention.

[0055] Next, the 3D point cloud data collected in step S110 is simultaneously input into the 3D point cloud target detection algorithm unit and the 3D point cloud segmentation and clustering algorithm unit. Specifically:

[0056] In step S120, the 3D point cloud data is input into a 3D point cloud object detection algorithm unit, and the 3D point cloud object detection algorithm unit determines an output result of a 3D detection bounding box of the detection object.

[0057] The 3D point cloud target detection algorithm unit includes a preprocessing module, a feature extraction module, an aggregation and bounding box prediction module, and a post-processing module. It predicts the bounding box information of the target by using voxel-based feature extraction and aggregation methods.

[0058] More specifically, although not shown in the figure, the 3D point cloud target detection algorithm unit further determines the information of the bounding box of the detected target through the following steps:

[0059] CXWL2022015A-A57

[0060] (1) Preprocessing: The original 3D point cloud data is divided into regions of interest and converted into three-dimensional voxels.

[0061] In the embodiment of the present invention, the range of the region of interest is set to 0 to 100 m in front of the radar, 50 to 50 m to the left and right, and 2 to 4 m up and down, and the three-dimensional voxel size is set to 10 cm × 10 cm × 15 cm, so the total number of voxels is 1000 × 1000 × 40.

[0062] Since the point cloud is unevenly distributed in space and there are many empty voxels, only the voxels with points in the voxels are retained as output valid voxels in the calculation. Each valid voxel can be identified by its voxel coordinates, and the features in the voxel can be averaged according to the coordinates of the original 3D point cloud data in the voxel.

[0063] By preprocessing the original 3D point cloud data (N1*4, N2*4, N3*4), sparse voxel features (M1*4, M2*4, M3*4) are output.

[0064] (2) Feature extraction: 3D and 2D feature extraction is performed on point cloud voxels at different times. The sparse voxel features are extracted in 3D by using three-dimensional sparse convolution (3DSparseConv) and scattered into the 2D bird's-eye view BEV dense voxels.

[0065] Specifically, the sparse voxel features (M1*4, M2*4, M3*4) are input into the feature extraction module, and the output is unified into BEV dense voxel features 1*256*125*125. Then, the BEV voxel features are deep-extracted through the two-dimensional convolution (2DConv) feature extraction module, and a unified deep feature 1*512*32*32 is output.

[0066] (3) Feature aggregation and bounding box prediction: The deep features of three adjacent frames are aggregated. Feature aggregation refers to feature splicing in the feature dimension. The unified deep features 1*512*32*32 of the three adjacent frames are aggregated to obtain the aggregated features 1*1536*32*32, and the feature level is fused through the convolution layer. Finally, the prediction layer head is used to predict the orientation, size offset, position offset, category and score of the target bounding box.

[0067] (4) Post-processing: Decode the output bounding box information and perform non-maximum suppression (NMS) to obtain the three-dimensional bounding box information of the detected target in the scene, that is, the 3D detection bounding box information.

[0068] In the embodiment of the present invention, the region of interest refers to the scene range where the target to be detected is located, and can be set according to actual needs, but this does not constitute a limitation to the present invention.

[0069] In step S130, the 3D point cloud data is input into a 3D point cloud segmentation and clustering algorithm unit, and the 3D point cloud segmentation and clustering algorithm unit determines an output result of a 3D clustering bounding box of the detection target.

[0070] The 3D point cloud segmentation and clustering algorithm unit extracts high-dimensional features from point clouds in three consecutive frames through the feature extraction module, feature aggregation and semantic segmentation module, point cloud ground point filtering module, and point cloud clustering algorithm.

[0071] More specifically, the original 3D point cloud data (N1*4, N2*4, N3*4) is input into the feature extraction module CXWL2022015A-A57

[0072] Block, each frame of point cloud is fixedly sampled to M*4 within the range of interest, and feature extraction is performed on the sampled point cloud to obtain deep features (M*1024, M*1024, M*1024).

[0073] The feature aggregation and semantic segmentation module first uses the current frame point cloud features as the point features to be queried, performs neighborhood queries and grouping on the previous two frames of point clouds, aggregates the features of the previous two frames of point clouds into the current point cloud, and performs convolution and maximization operations, finally obtaining the current frame point cloud features (M*1024). The prediction layer head performs target semantic category and score prediction, and outputs the labels and confidences of different categories of point clouds.

[0074] The point cloud ground point filtering module is used to filter out the background point clouds classified as ground points and other non-ground points, and only the foreground point clouds above the ground are retained.

[0075] The point cloud clustering algorithm refers to a search iterative strategy to aggregate point clouds of the same category and locally clustered and obtain their external three-dimensional bounding boxes, and finally obtain the three-dimensional bounding box information of all detected targets in the scene, that is, the 3D clustering bounding box information.

[0076] Next, go to step S140.

[0077] In step S140, the output result of the 3D point cloud target detection algorithm unit and the output result of the 3D point cloud segmentation and clustering algorithm unit are fused, and the output result of the fusion information is determined.

[0078] In an embodiment of the present invention, when the 3D point cloud target detection algorithm unit decides to output a 3D detection bounding box of the detection target, and the 3D point cloud segmentation and clustering algorithm unit decides to output a 3D clustering bounding box of the detection target, the 3D overlap of the 3D detection bounding box and the 3D clustering bounding box is calculated, and based on a comparison between the 3D overlap and a preset overlap threshold, the output result of the fusion information is determined.

[0079] Further, in an embodiment of the present invention, when the 3D overlap is greater than a preset overlap threshold, the output result of the fusion information is the 3D detection bounding box; when the 3D overlap is less than the preset overlap threshold, the Euclidean distance between the 3D detection bounding box and the 3D clustering bounding box is calculated, wherein, when the Euclidean distance is less than a preset Euclidean distance threshold and the confidence of the 3D detection bounding box is greater than a preset confidence threshold, the output result of the fusion information is the 3D detection bounding box, and when the Euclidean distance is greater than the preset Euclidean distance threshold, the output result of the fusion information is the 3D detection bounding box.

[0080] Further, in an embodiment of the present invention, when the 3D point cloud target detection algorithm unit decides not to output the 3D detection bounding box of the detection target, and the 3D point cloud segmentation and clustering algorithm unit decides to output the 3D clustering bounding box of the detection target, the number of points of the detection target of the 3D clustering bounding box is judged, wherein, when the number of points is greater than a preset point threshold, the detection target is judged to be an obstacle; when the CXWL2022015A-A57

[0081] When the number of points is less than a preset point threshold, the detection target is judged to be a noise point or a misidentified obstacle.

[0082] In an embodiment of the present invention, the method further includes: classifying the detection target according to a preset classification algorithm, and the classification includes classified obstacles and unclassified obstacles.

[0083] In an embodiment of the present invention, the output result of the fusion information, i.e., the obstacle target information of the point cloud scene, includes the separable categories of the obstacles, and the output is the separable categories of the obstacles, the three-dimensional coordinates (x, y, z) in the radar coordinate system, the three-dimensional dimensions dx, dy, dz corresponding to the coordinate axes in the radar coordinate system, and their orientation angles, a total of 8-dimensional information.

[0084] More specifically, the output results of the 3D point cloud target detection algorithm unit and the output results of the 3D point cloud segmentation and clustering algorithm unit are fused through association matching, filtering and merging, and finally the three-dimensional bounding box information of the obstacles in the scene is obtained, that is, the regular obstacle and irregular obstacle detection boxes are merged and fused, and the final output is used as the target detection box of the detection scene.

[0085] The association matching mainly adopts the matching rules based on intersection over union (IOU) and Euclidean distance, and calculates the 3D IOU of the 3D detection bounding box and the 3D clustering bounding box identified in each frame respectively. When the 3D IOU is greater than the preset threshold, the 3D detection bounding box is output as the target bounding box; when the 3D IOU threshold is less than the preset threshold, the Euclidean distance between the 3D detection bounding box and the 3D clustering bounding box is calculated. When the Euclidean distance between the two bounding boxes is less than the preset threshold and the confidence is greater than the preset confidence threshold, the 3D detection bounding box is output as the target bounding box; when the Euclidean distance between the two bounding boxes is greater than the preset threshold, the 3D detection bounding box is considered unreliable, and the 3D clustering bounding box is output as the target bounding box.

[0086] In an embodiment of the present invention, the IOU and confidence threshold settings are different for different detection targets. For example, for larger targets (such as vehicles), the IOU threshold is set to 0.5, and the confidence threshold is set to 0.3. For smaller targets (such as pedestrians or other small obstacles), the IOU threshold is set to 0.3, the confidence threshold is set to 0.3, and the Euclidean distance threshold is set to 0.1m.

[0087] Filtering is mainly used when the 3D point cloud target detection algorithm unit does not recognize the 3D detection bounding box, that is, does not output the 3D detection bounding box, but the 3D point cloud segmentation and clustering algorithm unit generates a 3D clustering bounding box. At this time, it is necessary to consider whether the point cloud point count of the target obstacle exceeds the preset threshold. When the point cloud point count of the target obstacle is greater than the preset threshold, it is considered to be an irregular other obstacle. When the point cloud point count of the target obstacle is less than the preset threshold, it is considered to be a noise point or a misidentified obstacle. For example, when the point cloud point count threshold is set to 10 points, if the point cloud point count of the target obstacle is greater than 10 points, the target is considered to be an obstacle. If the point cloud point count of the target obstacle is less than 10 points, the target is considered to be a noise point or a misidentified obstacle.

[0088] Merging refers to simply merging the retained 3D detected obstacles and 3D clustered obstacles as the obstacles for final scene recognition.

[0089] CXWL2022015A-A57

[0090] In the embodiment of the present invention, obstacles are divided into two categories: one is regular obstacles with clear classification, i.e., classified obstacles; the other is irregular obstacles whose categories cannot be determined, i.e., unclassified obstacles, which can be used for subsequent target tracking and fusion with other sensors.

[0091] Figure 2 A schematic diagram of a method for detecting a target at time t based on a solid-state laser radar through the method described in steps S110 to S140 is shown as an example.

[0092] like Figure 2 As shown, the point cloud data of the three consecutive frames of point cloud data in the time series, the point cloud data input at time t-2, the point cloud data input at time t-1, and the point cloud data input at time t, are processed by the 3D point cloud target detection algorithm unit and the 3D point cloud segmentation and clustering algorithm unit respectively. The related processing of the 3D point cloud target detection algorithm unit and the 3D point cloud segmentation and clustering algorithm unit is as mentioned above. Figure 1 In order not to cause unnecessary ambiguity to the present invention, the Figure 2 Description.

[0093] Figure 3 The system 300 for sensor data fusion provided by the present invention is shown, including an acquisition module 301 , a detection module 302 , a clustering module 303 and a fusion module 304 .

[0094] The acquisition module 301 is used to acquire 3D point cloud data within a predetermined detection range acquired by an acquisition device.

[0095] The detection module 302 is used to input the 3D point cloud data into a 3D point cloud target detection algorithm unit, and the 3D point cloud target detection algorithm unit determines the output result of the 3D detection bounding box of the detection target.

[0096] The clustering module 303 is used to input the 3D point cloud data into a 3D point cloud segmentation clustering algorithm unit, and the 3D point cloud segmentation clustering algorithm unit determines the output result of the 3D clustering bounding box of the detection target.

[0097] The fusion module 304 is used to fuse the output result of the 3D point cloud target detection algorithm unit and the output result of the 3D point cloud segmentation and clustering algorithm unit, and determine the output result of the fusion information.

[0098] In the embodiment of the present invention, the system further includes: a classification module 305. The classification module 305 is used to classify the detection target according to a preset classification algorithm, and the classification includes classified obstacles and unclassified obstacles.

[0099] The perception scheme of the present invention can significantly improve the perception performance of solid-state laser radar, and has important application value for actual roadside and vehicle-mounted solid-state laser radar perception schemes.

[0100] See also Figure 4 The present disclosure also provides an electronic device 40, which includes:

[0101] at least one processor; and,

[0102] a memory communicatively connected to the at least one processor; wherein,

[0103] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the perception method for target detection in the aforementioned method embodiment.

[0104] The present disclosure also provides a non-transitory computer-readable storage medium.

[0105] The readable storage medium stores computer instructions, which are used to enable the computer to execute the perception method for target detection in the aforementioned method embodiment.

[0106] The embodiments of the present disclosure also provide a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the perception method for target detection in the aforementioned method embodiment.

[0107] Reference below Figure 4 , which shows a schematic diagram of the structure of an electronic device 40 suitable for implementing the embodiment of the present disclosure. The electronic device in the embodiment of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0108] like Figure 4 As shown, the electronic device 40 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage device 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the electronic device 40 are also stored. The processing device 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0109] Typically, the following devices may be connected to the I / O interface 405: input devices 406 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 407 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 408 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 409. The communication device 409 may allow the electronic device 40 to communicate wirelessly or wired with other devices to exchange data. Although the electronic device 40 with various devices is shown in the figure, it should be understood that it is not required to implement or have all the devices shown. More or fewer devices may be implemented or have instead.

[0110] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 409, or installed from the storage device 408, or installed from the ROM 402. When the computer program is executed by the processing device 401, the above CXWL2022015A-A57 defined in the method of the embodiment of the present disclosure is executed.

[0111] Described function.

[0112] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0113] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0114] The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains at least two Internet Protocol addresses; sends a node evaluation request including the at least two Internet Protocol addresses to a node evaluation device, wherein the node evaluation device selects an Internet Protocol address from the at least two Internet Protocol addresses and returns it; receives the Internet Protocol address returned by the node evaluation device; wherein the obtained Internet Protocol address indicates an edge node in a content distribution network.

[0115] Alternatively, the computer-readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device: receives a node evaluation request including at least two Internet Protocol addresses; selects an Internet Protocol address from the at least two Internet Protocol addresses; and returns the selected Internet Protocol address; wherein the received Internet Protocol address indicates an edge node in a content distribution network.

[0116] CXWL2022015A-A57

[0117] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate 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 a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0118] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0119] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware. The name of a unit does not limit the unit itself in some cases. For example, the first acquisition unit may also be described as a "unit for acquiring at least two Internet Protocol addresses".

[0120] It should be understood that various parts of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.

[0121] The above is only a specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present disclosure should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A perception method for target detection, characterized in that: The method comprises: Acquire 3D point cloud data within a predetermined detection range collected by a collection device; Inputting the 3D point cloud data into a 3D point cloud target detection algorithm unit, and the 3D point cloud target detection algorithm unit determines an output result of a 3D detection bounding box of the detection target; Inputting the 3D point cloud data into a 3D point cloud segmentation and clustering algorithm unit, and allowing the 3D point cloud segmentation and clustering algorithm unit to determine an output result of a 3D clustering bounding box of a detection target; and The output result of the 3D point cloud target detection algorithm unit and the output result of the 3D point cloud segmentation and clustering algorithm unit are fused, and the output result of the fusion information is determined.

2. The perception method for target detection according to claim 1, characterized in that: When the 3D point cloud target detection algorithm unit decides to output the 3D detection bounding box of the detection target, and the 3D point cloud segmentation and clustering algorithm unit decides to output the 3D clustering bounding box of the detection target, the 3D overlap of the 3D detection bounding box and the 3D clustering bounding box is calculated, and based on the comparison between the 3D overlap and a preset overlap threshold, the output result of the fusion information is determined.

3. The perception method for target detection according to claim 2, characterized in that: When the 3D overlap is greater than a preset overlap threshold, the output result of the fusion information is the 3D detection bounding box; as well as When the 3D overlap is less than a preset overlap threshold, the Euclidean distance between the 3D detection bounding box and the 3D clustering bounding box is calculated, wherein: When the Euclidean distance is less than a preset Euclidean distance threshold, and the confidence of the 3D detection bounding box is greater than a preset confidence threshold, the output result of the fusion information is the 3D detection bounding box, and When the Euclidean distance is greater than a preset Euclidean distance threshold, the output result of the fusion information is the 3D clustering bounding box.

4. The perception method for target detection according to claim 1, characterized in that: The method further comprises: When the 3D point cloud target detection algorithm unit decides not to output the 3D detection bounding box of the detected target, CXWL2022015A-A57 When the 3D point cloud segmentation clustering algorithm unit decides to output the 3D clustering bounding box of the detection target, the number of points of the detection target in the 3D clustering bounding box is judged, wherein: When the number of points is greater than a preset point threshold, determining that the detected target is an obstacle; and When the number of points is less than a preset point threshold, the detection target is determined to be a noise point or a misidentified obstacle.

5. The perception method for target detection according to claim 4, characterized in that: The method further comprises: The detection target is classified according to a preset classification algorithm, and the classification includes classified obstacles and unclassified obstacles.

6. A perception system for target detection, characterized in that: The system comprises: An acquisition module is configured to acquire 3D point cloud data within a predetermined detection range acquired by an acquisition device; A detection module, configured to input the 3D point cloud data into a 3D point cloud object detection algorithm unit, and the 3D point cloud object detection algorithm unit determines an output result of a 3D detection bounding box of a detection target; a clustering module configured to input the 3D point cloud data into a 3D point cloud segmentation clustering algorithm unit, and the 3D point cloud segmentation clustering algorithm unit determines an output result of a 3D clustering bounding box of a detection target; and The fusion module is configured to fuse the output result of the 3D point cloud target detection algorithm unit and the output result of the 3D point cloud segmentation and clustering algorithm unit, and determine the output result of the fusion information.

7. The perception system for target detection according to claim 6, characterized in that: The system further comprises: The classification module is configured to classify the detection target according to a preset classification algorithm, and the classification includes classified obstacles and unclassified obstacles.

8. An electronic device, comprising: at least one processor; as well as, A memory communicatively connected to the at least one processor; wherein CXWL2022015A-A57 The memory stores instructions that can be executed by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor executes the perception method for target detection as described in any one of claims 1 to 5.

9. A non-transitory computer-readable storage medium storing computer instructions which, when executed by at least one processor, cause the at least one processor to perform a perception method for target detection as described in any one of claims 1 to 5.

10. A computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises program instructions, which, when executed by a computer, cause the computer to execute a perception method for target detection as described in any one of claims 1 to 5.

Images