Obstacle detection method and device, storage medium and electronic device

Abstract

The invention discloses an obstacle detection method and device, a storage medium and an electronic device, and relates to the technical field of vehicles. The method comprises the following steps: acquiring first sensing data and second sensing data of a target vehicle; semantic features of the target obstacle are extracted from the first sensing data, and structural features of the target obstacle are extracted from the second sensing data; performing feature fusion on the semantic features and the structural features based on the target corresponding relationship to obtain fused features; and detecting the target obstacle based on the fusion feature to obtain a detection result. According to the invention, the technical problem that the obstacle type cannot be correctly identified and the obstacle position cannot be correctly calculated due to the influence of a complex environment in a complex and changeable road environment in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle technology, and in particular to an obstacle detection method, device, storage medium and electronic device. Background Art

[0002] In the field of autonomous driving, obstacle detection is one of the key technologies to achieve safe driving. Correctly identifying obstacle types and calculating obstacle positions are conducive to ensuring the safety of unmanned driving. At present, the mainstream obstacle detection methods mainly include two categories: visual image-based processing and laser point cloud-based processing. However, these two methods each have significant limitations.

[0003] Based on the processing of visual images, the obstacle detection is easily affected by lighting conditions. For example, at night, in strong light or shadow environment, the image quality is reduced, resulting in reduced detection accuracy. Especially in complex three-dimensional environments, due to the lack of depth information, it is difficult to accurately judge the distance and three-dimensional shape of obstacles. Based on the processing of laser point clouds, when detecting obstacles, although the distance can be accurately measured, it is difficult to distinguish obstacles in environments lacking texture (such as white walls and smooth surfaces). Moreover, there is only point cloud structure position information, and the ability to distinguish the semantic types of obstacles is insufficient. Summary of the invention

[0004] The embodiments of the present invention provide an obstacle detection method, device, storage medium and electronic device to at least solve the technical problem that the related art cannot correctly identify the obstacle type and calculate the obstacle position in a complex and changeable road environment due to the influence of the complex environment.

[0005] According to one embodiment of the present invention, there is provided an obstacle detection method, comprising: acquiring first perception data and second perception data of a target vehicle, wherein the first perception data is image data of the target obstacle acquired by a camera, the second perception data is point cloud data of the target obstacle acquired by a laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle; extracting semantic features of the target obstacle from the first perception data, and extracting structural features of the target obstacle from the second perception data; performing feature fusion on the semantic features and the structural features based on a target correspondence relationship to obtain a fused feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data; and detecting the target obstacle based on the fused feature to obtain a detection result.

[0006] Optionally, obtaining the second perception data includes: obtaining third perception data of the target obstacle collected by the laser radar at a first collection moment; performing point cloud compensation on the third perception data based on the first collection moment, the second collection moment and the point cloud collection frequency to obtain second perception data, wherein the point cloud collection frequency is the collection frequency of the laser radar for collecting the third perception data, the second collection moment is the moment when the camera collects the first perception data, and the second perception data is aligned with the perception data of the first perception data at the second collection moment.

[0007] Optionally, feature fusion of semantic features and structural features is performed based on the target correspondence relationship to obtain fused features, including: determining the relative position of first position information and second position information, wherein the first position information is the installation position information of the camera, and the second position information is the installation position information of the lidar; based on the relative position, projecting the second perception data onto a normalized plane to obtain the target correspondence relationship; feature fusion of semantic features and structural features is performed based on a preset fusion algorithm and the target correspondence relationship to obtain a fused feature.

[0008] Optionally, the method also includes: planning an obstacle avoidance path of the target vehicle based on the detection results, a preset planning algorithm, the driving scenario and driving requirements of the target vehicle; and controlling the target vehicle to travel based on the obstacle avoidance path.

[0009] Optionally, extracting semantic features of the target obstacle from the first perception data includes: performing feature extraction on the first perception data based on a preset image segmentation algorithm to obtain semantic features, wherein the semantic features include obstacle type features and obstacle color features.

[0010] Optionally, extracting structural features of the target obstacle from the second perception data includes: performing feature extraction on the second perception data based on a preset point cloud segmentation algorithm to obtain structural features, wherein the structural features include obstacle position features and obstacle shape features.

[0011] Optionally, before extracting the semantic features of the target obstacle from the first perception data and extracting the structural features of the target obstacle from the second perception data, the method also includes: performing data denoising and data enhancement processing on the first perception data and the second perception data.

[0012] Optionally, the method further includes: displaying the detection result on a display of the target vehicle.

[0013] Optionally, cameras are installed on the front, rear, left and right sides of the target vehicle, and a lidar is installed on the top of the target vehicle.

[0014] According to one embodiment of the present invention, there is also provided an obstacle detection device, including: an acquisition module, used to acquire first perception data and second perception data of a target vehicle, wherein the first perception data is image data of the target obstacle collected by a camera, and the second perception data is point cloud data of the target obstacle collected by a lidar, and the target obstacle is an obstacle within a preset range of the target vehicle; an extraction module, used to extract semantic features of the target obstacle from the first perception data, and to extract structural features of the target obstacle from the second perception data; a fusion module, used to perform feature fusion of semantic features and structural features based on a target correspondence relationship to obtain a fused feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data; and a detection module, used to detect the target obstacle based on the fused feature to obtain a detection result.

[0015] Optionally, the acquisition module is also used to acquire third perception data of the target obstacle collected by the laser radar at the first collection moment; perform point cloud compensation on the third perception data based on the first collection moment, the second collection moment and the point cloud collection frequency to obtain second perception data, wherein the point cloud collection frequency is the collection frequency of the laser radar for collecting the third perception data, the second collection moment is the moment when the camera collects the first perception data, and the second perception data is aligned with the perception data of the first perception data at the second collection moment.

[0016] Optionally, the fusion module is also used to determine the relative position of the first position information and the second position information, wherein the first position information is the installation position information of the camera, and the second position information is the installation position information of the lidar; based on the relative position, the second perception data is projected onto a normalized plane to obtain a target correspondence relationship; based on a preset fusion algorithm and the target correspondence relationship, feature fusion is performed on the semantic features and the structural features to obtain a fused feature.

[0017] Optionally, the extraction module is further used to perform feature extraction on the first perception data based on a preset image segmentation algorithm to obtain semantic features, wherein the semantic features include obstacle type features and obstacle color features.

[0018] Optionally, the extraction module is also used to perform feature extraction on the second perception data based on a preset point cloud segmentation algorithm to obtain structural features, wherein the structural features include obstacle position features and obstacle shape features.

[0019] Optionally, the extraction module is also used to perform data denoising and data enhancement processing on the first perception data and the second perception data.

[0020] Optionally, the detection module is further used to display the detection result on a display of the target vehicle.

[0021] According to one embodiment of the present invention, a vehicle is provided, and the vehicle is used to execute any of the above obstacle detection methods.

[0022] According to one embodiment of the present invention, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute any of the above obstacle detection methods when running on a computer or a processor.

[0023] According to one embodiment of the present invention, there is further provided an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the above obstacle detection methods.

[0024] According to one embodiment of the present invention, a computer program product is further provided, including a computer program, and when the computer program is executed by a processor, the obstacle detection method in any one of the above items is implemented.

[0025] In an embodiment of the present invention, by acquiring first perception data and second perception data of a target vehicle, wherein the first perception data is image data of a target obstacle collected by a camera, the second perception data is point cloud data of a target obstacle collected by a lidar, and the target obstacle is an obstacle within a preset range of the target vehicle; extracting semantic features of the target obstacle from the first perception data, and extracting structural features of the target obstacle from the second perception data; performing feature fusion on semantic features and structural features based on target correspondence to obtain fused features, wherein the target correspondence is an association between the first perception data and the second perception data; detecting the target obstacle based on the fused features to obtain a detection result, the purpose of effectively fusing the data of the two sensors and achieving complementary advantages is achieved, thereby achieving the technical effect of improving the accuracy, robustness and environmental adaptability of obstacle detection, especially in a complex and changeable road environment, ensuring the safety and reliability of the autonomous driving system, and thus solving the technical problem that the related technology cannot correctly identify the obstacle type and calculate the obstacle position in a complex and changeable road environment due to the influence of the complex environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0027] Figure 1 is a flow chart of an obstacle detection method according to one embodiment of the present invention;

[0028] Figure 2 is a structural block diagram of an obstacle detection device according to one embodiment of the present invention. DETAILED DESCRIPTION

[0029] To facilitate understanding, some descriptions of concepts related to the embodiments of the present invention are exemplarily provided for reference.

[0030] As shown below:

[0031] Visual sensor (camera): installed on the front, rear, left and right sides of the vehicle, using a wide-angle lens to cover a wider field of view, and the number of installed cameras is 4 or more.

[0032] Laser Radar (LiDAR): Installed on the top of the vehicle, it ensures unobstructed scanning of the surrounding environment and generates 360-degree point cloud data.

[0033] Data processing unit (including GPU): Located inside the vehicle, responsible for receiving and processing data from various sensors and executing obstacle detection algorithms.

[0034] Controller: Plans the obstacle avoidance path based on the obstacle fusion results and controls the vehicle's movement.

[0035] Display: Based on the obstacle fusion results, it is displayed on the human-computer interaction page to facilitate the driver to observe the surrounding obstacle information.

[0036] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In the description of this embodiment, unless otherwise specified, the meaning of "multiple" is two or more. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] According to one embodiment of the present invention, an embodiment of an obstacle detection method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] The method embodiment can be executed in an electronic device including a memory and a processor, a similar control device or a system. Taking an electronic device as an example, the electronic device may include one or more processors and a memory for storing data. Optionally, the electronic device may also include a communication device and a display device for communication functions. It can be understood by a person of ordinary skill in the art that the above structural description is only illustrative and does not limit the structure of the above electronic device. For example, the electronic device may also include more or fewer components than the above structural description, or have a configuration different from the above structural description.

[0040] The processor may include one or more processing units. For example, the processor may include a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processing (DSP) chip, a microcontroller unit (MCU), a field-programmable gate array (FPGA), a neural network processor (NPU), a tensor processing unit (TPU), an artificial intelligence (AI) type processor, and other processing devices. Among them, different processing units may be independent components or integrated into one or more processors. In some instances, the electronic device may also include one or more processors.

[0041] The memory can be used to store computer programs, such as storing the computer program corresponding to the obstacle detection method in the embodiment of the present invention. The processor implements the above-mentioned obstacle detection method by running the computer program stored in the memory. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory may further include a memory remotely arranged relative to the processor, and these remote memories can be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0042] The communication device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the communication device includes a network adapter (network interface controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the communication device can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0043] The display device may be, for example, a touch screen type liquid crystal display (LCD) and a touch display (also referred to as a "touch screen" or "touch display screen"). The liquid crystal display may enable a user to interact with a user interface of a mobile terminal. In some embodiments, the mobile terminal has a graphical user interface (GUI), and a user may interact with the GUI by finger contact and / or gestures on a touch-sensitive surface, wherein the human-computer interaction functions here may optionally include the following interactions: creating web pages, drawing, word processing, making electronic documents, games, video conferencing, instant messaging, sending and receiving emails, call interfaces, playing digital videos, playing digital music and / or web browsing, etc. The executable instructions for executing the above human-computer interaction functions are configured / stored in a computer program product or a readable storage medium executable by one or more processors.

[0044] In this embodiment, an obstacle detection method running on an electronic device is provided. Figure 1 is a flow chart of an obstacle detection method according to one embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0045] Step S10, obtaining first perception data and second perception data of the target vehicle, wherein the first perception data is image data of the target obstacle collected by the camera, the second perception data is point cloud data of the target obstacle collected by the laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle;

[0046] In the embodiment of the present invention, the first perception data can be understood as image data of obstacles within a preset range collected by a camera on the target vehicle. Exemplarily, the first perception data can be a red green blue (RGB) image, which is not limited here.

[0047] The second perception data can be understood as the point cloud data of obstacles within a preset range collected by the laser radar on the target vehicle. For example, the laser radar installed on the top of the vehicle can scan the surrounding environment without obstruction, and the 360-degree point cloud data generated is the second perception data, which is not limited here.

[0048] Acquiring the first perception data and the second perception data of the target vehicle can be understood as collecting the image data and point cloud data of the target obstacle through the camera and the laser radar on the target vehicle. For example, the RGB image of the target obstacle within a preset range can be captured by the camera, and the point cloud data within the preset range can be collected by the laser radar, which is not limited here.

[0049] In an embodiment of the present invention, by acquiring first perception data and second perception data of the target vehicle, wherein the first perception data is image data of the target obstacle collected by the camera, and the second perception data is point cloud data of the target obstacle collected by the lidar, and the target obstacle is an obstacle within a preset range of the target vehicle, it is ensured that the layout of the camera and the lidar can reasonably and effectively cover all directions around the vehicle, thereby providing high-quality data input for the subsequent fusion algorithm.

[0050] Step S11, extracting semantic features of the target obstacle from the first perception data, and extracting structural features of the target obstacle from the second perception data;

[0051] In the embodiment of the present invention, the semantic features can be understood as some descriptive features in the image data of the target obstacle. Exemplarily, the semantic features can include the color, material, function and other features of the target obstacle, which are not limited here.

[0052] The structural features can be understood as some structural features in the image data of the target obstacle. For example, the structural features may include features such as the size, shape, and position of the target obstacle, which are not limited here.

[0053] Extracting the semantic features of the target obstacle from the first perception data can be understood as extracting the descriptive features of the target obstacle from the image data of the obstacle within the preset range collected by the camera. For example, the color, material, function and other features of the target obstacle can be extracted from the first perception data, which are not limited here.

[0054] Extracting the structural features of the target obstacle from the second perception data can be understood as extracting the structural features of the target obstacle from the point cloud data of the obstacle within the preset range collected by the laser radar. For example, the size, shape, position and other features of the target obstacle can be extracted from the second perception data, which are not limited here.

[0055] In an embodiment of the present invention, by extracting the semantic features of the target obstacle from the first perception data and extracting the structural features of the target obstacle from the second perception data, it can help to better understand and describe the shape and location of the target obstacle and identify it.

[0056] Step S12, performing feature fusion on the semantic feature and the structural feature based on the target correspondence relationship to obtain a fusion feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data;

[0057] In the embodiment of the present invention, the target correspondence relationship can be understood as the association relationship between the image data collected by the camera and the point cloud data collected by the laser radar. For example, the target correspondence relationship can be obtained by analyzing the installation position information of the camera and the installation position information of the laser radar, which is not limited here.

[0058] Feature fusion can be understood as fusing the semantic features and structural features of the target obstacle. For example, the color information in the semantic features of the target obstacle and the size information in the structural features can be fused, which is not limited here.

[0059] The semantic features and structural features are fused based on the target correspondence to obtain the fused features. It can be understood that the semantic features and structural features of the target obstacle are fused based on the correlation between the image data collected by the camera and the point cloud data collected by the lidar to obtain the fused features. Exemplarily, the semantic features and structural features of the target obstacle can be fused based on a preset fusion algorithm and the target correspondence to obtain the fused features, which is not limited here.

[0060] In an embodiment of the present invention, semantic features and structural features are fused based on target correspondence to obtain fused features, wherein the target correspondence is the association between the first perception data and the second perception data. By fusing data from different sources, information loss can be reduced, and the complementarity between data can be better utilized to improve the accuracy of the detection model.

[0061] Step S13: Detect the target obstacle based on the fusion features to obtain a detection result.

[0062] In the embodiment of the present invention, the detection result may be understood as the position information and distance information of the detected target obstacle from the target vehicle, etc., which is not limited here.

[0063] Detecting the target obstacle based on the fused features to obtain the detection result can be understood as detecting the target obstacle based on the fused semantic features and structural features, thereby giving the detection result of the target obstacle. For example, the target vehicle can plan an obstacle avoidance path based on the detection result and control the vehicle to travel, which is not limited here.

[0064] In the embodiment of the present invention, the target obstacle is detected based on the fusion feature to obtain the detection result, thereby achieving accurate detection of the obstacle and providing a more reliable obstacle avoidance strategy for the autonomous driving vehicle, thereby reducing the risk of accidents.

[0065] By acquiring the first perception data and the second perception data of the target vehicle, wherein the first perception data is the image data of the target obstacle collected by the camera, the second perception data is the point cloud data of the target obstacle collected by the laser radar, and the target obstacle is an obstacle within the preset range of the target vehicle, it is ensured that the layout of the camera and the laser radar can reasonably and effectively cover all directions around the vehicle, thereby providing high-quality data input for the subsequent fusion algorithm. By extracting the semantic features of the target obstacle from the first perception data and extracting the structural features of the target obstacle from the second perception data, it can help to better understand and describe the shape and location of the target obstacle and identify it. Based on the target correspondence relationship, the semantic features and the structural features are fused to obtain fusion features, wherein the target correspondence relationship is the association relationship between the first perception data and the second perception data. By fusing data from different sources, the loss of information can be reduced, and the complementarity between data can be better utilized to improve the accuracy of the detection model. The target obstacle is detected based on the fusion feature to obtain the detection result, which realizes the accurate detection of the obstacle, and provides a more reliable obstacle avoidance strategy for the autonomous driving vehicle, reducing the risk of accidents.

[0066] Through the above steps, the first perception data and the second perception data of the target vehicle are obtained, wherein the first perception data is the image data of the target obstacle collected by the camera, the second perception data is the point cloud data of the target obstacle collected by the laser radar, and the target obstacle is an obstacle within the preset range of the target vehicle; the semantic features of the target obstacle are extracted from the first perception data, and the structural features of the target obstacle are extracted from the second perception data; based on the target correspondence relationship, the semantic features and the structural features are feature fused to obtain the fused features, wherein the target correspondence relationship is the association relationship between the first perception data and the second perception data; based on the fused features, the target obstacle is detected to obtain the detection result, so as to achieve the purpose of effectively fusing the data of the two sensors and realizing the complementary advantages, thereby achieving the technical effect of improving the accuracy, robustness and environmental adaptability of obstacle detection, especially in a complex and changeable road environment, ensuring the safety and reliability of the automatic driving system, and thus solving the technical problem that the related technology cannot correctly identify the obstacle type and calculate the obstacle position in a complex and changeable road environment due to the influence of the complex environment.

[0067] Optionally, in step S10, obtaining the second perception data may include the following execution steps:

[0068] Step S101, obtaining third perception data of the target obstacle collected by the laser radar at the first collection moment;

[0069] Step S102, performing point cloud compensation on the third perception data based on the first acquisition moment, the second acquisition moment and the point cloud acquisition frequency to obtain second perception data, wherein the point cloud acquisition frequency is the acquisition frequency of the third perception data acquired by the laser radar, the second acquisition moment is the moment when the camera acquires the first perception data, and the second perception data is aligned with the perception data of the first perception data at the second acquisition moment.

[0070] In the embodiment of the present invention, the first collection moment can be understood as the moment when the laser radar collects the third perception data of the target obstacle.

[0071] The third perception data can be understood as the initial point cloud data of obstacles within a preset range collected by the lidar.

[0072] The second acquisition moment can be understood as the moment when the camera acquires image data of obstacles within a preset range.

[0073] The second perception data can be understood as the data obtained after point cloud compensation is performed on the initial point cloud data of obstacles within a preset range collected by the laser radar.

[0074] Acquiring the third perception data of the target obstacle collected by the laser radar at the first collection moment can be understood as acquiring the initial point cloud data of the target obstacle collected by the laser radar at the first collection moment.

[0075] Based on the first acquisition time, the second acquisition time and the point cloud acquisition frequency, point cloud compensation is performed on the third perception data to obtain the second perception data. This can be understood as performing point cloud compensation on the initial point cloud data of the target obstacle based on the time when the lidar collects the initial point cloud data of the target obstacle, the time when the camera collects the obstacle image data within a preset range, and the acquisition frequency of the lidar to collect the initial point cloud data of the target obstacle to obtain the second perception data.

[0076] In an embodiment of the present invention, by acquiring third perception data of a target obstacle collected by a laser radar at a first collection moment, point cloud compensation is performed on the third perception data based on the first collection moment, the second collection moment and the point cloud collection frequency, to obtain second perception data, wherein the point cloud collection frequency is the collection frequency of the third perception data collected by the laser radar, the second collection moment is the moment when the camera collects the first perception data, and the second perception data is aligned with the perception data of the first perception data at the second collection moment, thereby improving the detection accuracy of the target.

[0077] Optionally, in step S12, performing feature fusion on the semantic features and the structural features based on the target correspondence to obtain the fused features may include the following execution steps:

[0078] Step S121, determining the relative position of the first position information and the second position information, wherein the first position information is the installation position information of the camera, and the second position information is the installation position information of the laser radar;

[0079] Step S122, projecting the second perception data onto a normalized plane based on the relative position to obtain a target correspondence relationship;

[0080] Step S123, performing feature fusion on the semantic features and the structural features based on a preset fusion algorithm and the target correspondence relationship to obtain a fused feature.

[0081] In the embodiment of the present invention, the first position information can be understood as the position information of the camera in the target vehicle for collecting the image data of the target obstacle. For example, the camera is installed at the front, rear, left and right sides of the target vehicle, using a wide-angle lens to cover a wider field of view, and the number of cameras installed is 4 or more, which is not limited here.

[0082] The second position information can be understood as the position information of the laser radar in the target vehicle for collecting the point cloud data of the target obstacle. For example, the laser radar is installed on the top of the target vehicle to ensure unobstructed scanning of the surrounding environment and generate 360-degree point cloud data, which is not limited here.

[0083] The relative position can be understood as the positional relationship between the position information of the camera in the target vehicle and the position information of the lidar.

[0084] The installation position information can be understood as the rotation matrix and translation vector determined according to the position information of the camera in the target vehicle or the position information of the laser radar, which is not limited here.

[0085] The preset fusion algorithm can be understood as an algorithm for fusing semantic features with structural features, thereby enabling the complementarity of two data types in different feature contents, which is not limited here.

[0086] Determining the relative position of the first position information and the second position information can be understood as determining the relative position relationship between the camera and the laser radar based on the installation position information of the camera and the installation position information of the laser radar.

[0087] Projecting the second perception data onto the normalized plane based on the relative position to obtain the target correspondence relationship can be understood as projecting the obstacle point cloud onto the normalized plane of the visual image according to the relative position to obtain a one-to-one correspondence between the point cloud and the pixel, thereby realizing the association between the point cloud obstacle and the visual obstacle, which is not limited here.

[0088] Based on the preset fusion algorithm and the target correspondence, the semantic features and the structural features are fused to obtain the fused features, which can be understood as follows: according to the correlation between the point cloud obstacles and the visual obstacles, the fused obstacle features include the semantic features provided by the visual obstacles and the structural features provided by the point cloud obstacles, which are not limited here.

[0089] In an embodiment of the present invention, by determining the relative position of first position information and second position information, wherein the first position information is the installation position information of the camera, and the second position information is the installation position information of the laser radar, the second perception data is projected onto a normalized plane based on the relative position to obtain a target correspondence relationship, and based on a preset fusion algorithm and the target correspondence relationship, feature fusion of semantic features and structural features is performed to obtain fused features, thereby effectively improving the accuracy, stability and robustness of target detection, thereby enabling the system to perform better performance in practical applications.

[0090] Optionally, the obstacle detection method further includes:

[0091] Plan the target vehicle's obstacle avoidance path based on the detection results, preset planning algorithm, the target vehicle's driving scenario and driving requirements;

[0092] Control the target vehicle to travel based on the obstacle avoidance path.

[0093] In the embodiment of the present invention, the preset planning algorithm can be understood as an algorithm used to plan a route for the target vehicle to avoid obstacles. Exemplarily, the preset planning algorithm can be an obstacle avoidance strategy and a control algorithm, which is not limited here.

[0094] The driving scene can be understood as the driving environment that the target vehicle is currently in. For example, the driving scene may include a highway, a congested section of a city, night driving, bad weather, etc., which are not limited here.

[0095] Driving requirements can be understood as some requirements specified by the user for the target vehicle when it is in a driving situation. For example, driving requirements can include safety, comfort, reliability, etc., which are not limited here.

[0096] The obstacle avoidance path can be understood as the route that the vehicle takes to avoid the obstacle and continue driving when an obstacle is detected. It is not restricted here.

[0097] Planning the obstacle avoidance path of the target vehicle based on the detection results, the preset planning algorithm, the driving scenario and driving requirements of the target vehicle can be understood as planning a feasible obstacle avoidance route for the target vehicle based on the detection results of the target obstacles, the preset algorithm for planning the obstacle avoidance route for the target vehicle, the driving scenario of the target vehicle and some requirements specified by the user for the target vehicle when in the driving situation, so that the target vehicle can continue to drive.

[0098] Controlling the target vehicle to travel based on the obstacle avoidance path can be understood as controlling the vehicle to continue traveling based on the planned obstacle avoidance route.

[0099] In an embodiment of the present invention, an obstacle avoidance path of a target vehicle is planned based on detection results, a preset planning algorithm, the driving scenario and driving requirements of the target vehicle, and then the target vehicle is controlled to travel based on the obstacle avoidance path, so that the target vehicle can be effectively prevented from colliding with obstacles or other vehicles, thereby reducing the risk of traffic accidents.

[0100] Optionally, in step S11, extracting the semantic features of the target obstacle from the first perception data may include the following execution steps:

[0101] Step S111, extracting features from the first perception data based on a preset image segmentation algorithm to obtain semantic features, wherein the semantic features include obstacle type features and obstacle color features.

[0102] In the embodiment of the present invention, the preset image segmentation algorithm can be understood as an algorithm for segmenting the image data collected by the camera, thereby obtaining the type and color information of each obstacle. Exemplarily, the image segmentation algorithm includes threshold segmentation, region growing, edge detection, etc., which are not limited here.

[0103] Based on the feature extraction of the first perception data, obtaining the semantic feature can be understood as extracting features from the image data using a preset image segmentation algorithm, thereby obtaining the type and color information of the target obstacle, etc. Exemplarily, the image data is processed by an image segmentation algorithm, deep learning, and other methods to obtain semantic features such as the type and color of the obstacle, which is not limited here.

[0104] In an embodiment of the present invention, feature extraction is performed on the first perception data based on a preset image segmentation algorithm to obtain semantic features, wherein the semantic features include obstacle type features and obstacle color features, which can more accurately identify and classify different obstacles, thereby improving the target vehicle's perception of the environment.

[0105] Optionally, in step S11, extracting the structural features of the target obstacle from the second perception data may include the following execution steps:

[0106] Step S112, extracting features from the second perception data to obtain structural features, wherein the structural features include obstacle position features and obstacle shape features.

[0107] In the embodiment of the present invention, the preset point cloud segmentation algorithm can be understood as an algorithm for segmenting the point cloud data collected by the laser radar, thereby obtaining the position characteristics and shape characteristics of each obstacle. Exemplarily, the point cloud segmentation algorithm includes a point cloud segmentation algorithm based on geometric features, a point cloud segmentation algorithm based on deep learning, etc., which are not limited here.

[0108] Based on the feature extraction of the second perception data, the structural features can be understood as extracting features from the point cloud data using a preset point cloud segmentation algorithm, thereby obtaining the position features and shape features of the target obstacle. Exemplarily, the point cloud data is processed by a point cloud segmentation algorithm to obtain structural features such as the position and shape of the obstacle, which is not limited here.

[0109] In an embodiment of the present invention, structural features are obtained based on feature extraction of the second perception data, wherein the structural features include obstacle position features and obstacle shape features. This can provide a more comprehensive understanding of the environment surrounding the target vehicle, improve the ability to perceive and understand obstacles, and thus better plan actions and make decisions.

[0110] Optionally, in step S11, before extracting the semantic features of the target obstacle from the first perception data and extracting the structural features of the target obstacle from the second perception data, the obstacle detection method further includes:

[0111] Data denoising and data enhancement are performed on the first perception data and the second perception data.

[0112] In the embodiment of the present invention, denoising can be understood as removing noise points from the initial image data and point cloud data. Exemplarily, the data denoising method may include median filtering, Gaussian filtering, mean filtering, edge preserving filtering, etc., which are not limited here.

[0113] Data enhancement processing can be understood as performing enhancement processing operations on the image data and point cloud data after denoising. Exemplarily, data enhancement processing methods may include rotation, flipping, scaling, translation, etc., which are not limited here.

[0114] Performing data denoising and data enhancement processing on the first perception data and the second perception data can be understood as performing data preprocessing operations on the image data collected by the camera and the point cloud data collected by the lidar, including denoising and data enhancement processing.

[0115] In an embodiment of the present invention, by performing data denoising and data enhancement processing on the first perception data and the second perception data, the quality and diversity of the data can be improved, thereby improving the performance and generalization ability of the detection model.

[0116] Optionally, the obstacle detection method further includes:

[0117] The detection results are displayed on the target vehicle's display.

[0118] In the embodiment of the present invention, displaying the detection result on the display of the target vehicle can be understood as the detection result of the target obstacle will be displayed on the display of the target vehicle. The detection result may include the size, color, position, etc. of the target obstacle, which is not limited here.

[0119] In the embodiment of the present invention, the detection result is displayed on the display of the target vehicle, thereby providing data support for the subsequent planning of the obstacle avoidance route, thereby improving the safety of the target vehicle during driving.

[0120] Optionally, the obstacle detection method further includes:

[0121] Cameras are installed on the front, rear, left and right sides of the target vehicle, and the lidar is installed on the top of the target vehicle.

[0122] In an embodiment of the present invention, cameras for collecting image data are installed on the front, rear, left and right sides of the target vehicle, and a laser radar for collecting point cloud data is installed on the top of the target vehicle, thereby effectively covering all directions around the vehicle and providing high-quality data input for subsequent fusion algorithms.

[0123] The technical solution proposed in this invention mainly includes seven core parts: data acquisition module, data preprocessing module, feature extraction module, data matching module, feature fusion module, decision-making, control, display, etc. By deeply fusing visual images with laser point cloud data, accurate recognition and classification of obstacles can be achieved.

[0124] The technical solution proposed in the present invention has the following steps: first, data collection and pre-processing operations are performed on the data, mainly including the camera capturing RGB images and the acquisition time, while the laser radar collects point cloud data and the acquisition time, the camera installation position (rotation matrix and translation vector), the laser radar installation position (rotation matrix and translation vector), and the visual image (i.e., RGB image and point cloud data) is denoised and enhanced. At the same time, according to the acquisition time of the image and point cloud and the point cloud acquisition frequency, the laser point cloud is compensated to the image acquisition time using the point cloud compensation algorithm, and then point cloud filtering, denoising and other processing are performed. Then, according to the camera installation position and the laser radar installation position, the relative position of the two is calculated. Secondly, feature extraction is performed on the processed image data and point cloud data, including extracting semantic features such as obstacle type and color from the image data through image segmentation algorithms, deep learning and other methods, and extracting structural features such as obstacle position and shape from the point cloud data through point cloud segmentation algorithms. Next, the image data and point cloud data are matched and feature fused. According to the relative position, the obstacle point cloud is projected onto the normalized plane of the visual image to obtain a one-to-one correspondence between the point cloud and the pixel, thereby realizing the association between the point cloud obstacle and the visual obstacle. Then, according to the association between the point cloud obstacle and the visual obstacle, the fused obstacle features include the semantic features provided by the visual obstacle and the structural features provided by the point cloud obstacle. Finally, according to the detection results of the target obstacle, a reasonable and effective obstacle avoidance path is planned to control the vehicle to continue driving.

[0125] The present invention proposes an efficient visual image and laser point cloud feature fusion algorithm, which can achieve the complementarity of the two data types in different feature contents, thereby improving the accuracy and robustness of obstacle detection. By designing a unique system structure design, including the design of data processing flow and module system structure, the efficiency and accuracy of data collection are improved, and false positives and missed positives are reduced. A data matching and feature fusion method for image and point cloud data is proposed, which can accurately extract the semantic features and structural features of obstacles, thereby providing new solutions and ideas for obstacle detection in the field of autonomous driving.

[0126] On the basis of maintaining the core concept of the present invention, in addition to using traditional RGB cameras, infrared cameras, depth cameras or event cameras can also be considered to enhance the perception ability in specific environments (such as low light, high dynamic range). According to the cost and application requirements, different types of LiDARs, such as solid-state LiDAR, can be selected, and their installation positions and quantities can be adjusted to achieve more flexible or more economical solutions. In the data preprocessing stage, in addition to basic denoising and enhancement, more advanced image processing and point cloud processing algorithms can be introduced, such as defogging algorithms based on deep learning, point cloud denoising algorithms, etc., to further improve data quality. In the feature extraction part, different deep learning models or point cloud segmentation algorithms can be tried to extract more diverse and representative features. In the feature fusion stage, in addition to the existing fusion algorithms, other fusion strategies can be explored, such as feature fusion based on attention mechanism, multimodal learning, etc., to further improve the fusion effect. In the decision and control part, more refined obstacle avoidance strategies and control algorithms can be designed according to different driving scenarios and requirements, such as path planning algorithms based on reinforcement learning, dynamic obstacle avoidance algorithms, etc. In terms of system integration, more efficient hardware architecture and parallel computing technology can be considered to accelerate the execution speed of the algorithm and improve the real-time performance of the system. Optimize sensor layout and data transmission protocol to reduce data transmission delay and loss, and improve system stability and reliability.

[0127] The present invention proposes for the first time an efficient algorithm for fusion of visual images and laser point cloud features. Especially in scenes with poor lighting or lack of texture information, the fused data can more accurately identify obstacles, realizing the complementarity of the two data types in different feature contents. The present invention avoids relying on visual detection of obstacle structural features and uses laser point clouds to detect the structural features of obstacles. Whether it is urban roads, rural roads or severe weather conditions, the system can maintain stable detection performance. Accurate detection of obstacles provides a more reliable obstacle avoidance strategy for autonomous driving vehicles and reduces the risk of accidents. By optimizing the system structure design and data processing flow, the efficiency and accuracy of data collection are improved, and false alarms and missed alarms are reduced. And through the fusion of the two types of data and necessary parallel computing, the real-time operation and high robustness of the algorithm in complex environments are ensured. The successful implementation of the present invention provides new ideas for technological innovation in the field of autonomous driving and promotes the improvement of industry standards.

[0128] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0129] In this embodiment, an obstacle detection device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0130] Figure 2 is a structural block diagram of an obstacle detection device according to one embodiment of the present invention. Figure 2 As shown, an obstacle detection device 200 is taken as an example, and the device includes: an acquisition module 201, which is used to acquire first perception data and second perception data of a target vehicle, wherein the first perception data is image data of the target obstacle collected by a camera, and the second perception data is point cloud data of the target obstacle collected by a laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle; an extraction module 202, which is used to extract semantic features of the target obstacle from the first perception data, and extract structural features of the target obstacle from the second perception data; a fusion module 203, which is used to perform feature fusion on semantic features and structural features based on a target correspondence relationship to obtain a fusion feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data; a detection module 204, which is used to detect the target obstacle based on the fusion feature to obtain a detection result.

[0131] Optionally, the acquisition module 201 is also used to acquire third perception data of the target obstacle collected by the laser radar at the first collection moment; perform point cloud compensation on the third perception data based on the first collection moment, the second collection moment and the point cloud collection frequency to obtain second perception data, wherein the point cloud collection frequency is the collection frequency of the laser radar for collecting the third perception data, the second collection moment is the moment when the camera collects the first perception data, and the second perception data is aligned with the perception data of the first perception data at the second collection moment.

[0132] Optionally, the fusion module 203 is also used to determine the relative position of the first position information and the second position information, wherein the first position information is the installation position information of the camera, and the second position information is the installation position information of the lidar; based on the relative position, the second perception data is projected onto a normalized plane to obtain a target correspondence relationship; based on a preset fusion algorithm and the target correspondence relationship, feature fusion is performed on the semantic features and the structural features to obtain a fused feature.

[0133] Optionally, the extraction module 202 is further used to perform feature extraction on the first perception data based on a preset image segmentation algorithm to obtain semantic features, wherein the semantic features include obstacle type features and obstacle color features.

[0134] Optionally, the extraction module 202 is further used to perform feature extraction on the second perception data based on a preset point cloud segmentation algorithm to obtain structural features, wherein the structural features include obstacle position features and obstacle shape features.

[0135] Optionally, the extraction module 202 is further used to perform data denoising and data enhancement processing on the first perception data and the second perception data.

[0136] Optionally, the detection module 204 is further configured to display the detection result on a display of the target vehicle.

[0137] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0138] An embodiment of the present invention further provides a vehicle, which is used to execute the steps in any of the above method embodiments.

[0139] An embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any of the above method embodiments when running on a computer or a processor.

[0140] Optionally, in this embodiment, the computer-readable storage medium may be configured to store a computer program for performing the following steps:

[0141] Step S10, obtaining first perception data and second perception data of the target vehicle, wherein the first perception data is image data of the target obstacle collected by the camera, the second perception data is point cloud data of the target obstacle collected by the laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle;

[0142] Step S11, extracting semantic features of the target obstacle from the first perception data, and extracting structural features of the target obstacle from the second perception data;

[0143] Step S12, performing feature fusion on the semantic feature and the structural feature based on the target correspondence relationship to obtain a fusion feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data;

[0144] Step S13: Detect the target obstacle based on the fusion features to obtain a detection result.

[0145] Optionally, in this embodiment, the above-mentioned computer-readable storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.

[0146] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0147] Optionally, in this embodiment, the processor in the electronic device may be configured to run a computer program to perform the following steps:

[0148] Step S10, obtaining first perception data and second perception data of the target vehicle, wherein the first perception data is image data of the target obstacle collected by the camera, the second perception data is point cloud data of the target obstacle collected by the laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle;

[0149] Step S11, extracting semantic features of the target obstacle from the first perception data, and extracting structural features of the target obstacle from the second perception data;

[0150] Step S12, performing feature fusion on the semantic feature and the structural feature based on the target correspondence relationship to obtain a fusion feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data;

[0151] Step S13: Detect the target obstacle based on the fusion features to obtain a detection result.

[0152] An embodiment of the present invention further provides a computer program product, including a computer program, which implements the steps in any of the above method embodiments when executed by a processor.

[0153] Optionally, in this embodiment, the computer program in the above computer program product may be configured to perform the following steps when executed by a processor:

[0154] Step S10, obtaining first perception data and second perception data of the target vehicle, wherein the first perception data is image data of the target obstacle collected by the camera, the second perception data is point cloud data of the target obstacle collected by the laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle;

[0155] Step S11, extracting semantic features of the target obstacle from the first perception data, and extracting structural features of the target obstacle from the second perception data;

[0156] Step S12, performing feature fusion on the semantic feature and the structural feature based on the target correspondence relationship to obtain a fusion feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data;

[0157] Step S13: Detect the target obstacle based on the fusion features to obtain a detection result.

[0158] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0159] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0160] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0161] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0162] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0163] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0164] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, magnetic disk or optical disk, etc., various media that can store program codes.

[0165] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An obstacle detection method, characterized in that: include: Acquire first perception data and second perception data of the target vehicle, wherein the first perception data is image data of the target obstacle collected by a camera, the second perception data is point cloud data of the target obstacle collected by a laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle; Extracting semantic features of the target obstacle from the first perception data, and extracting structural features of the target obstacle from the second perception data; Performing feature fusion on the semantic feature and the structural feature based on a target correspondence relationship to obtain a fusion feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data; The target obstacle is detected based on the fusion feature to obtain a detection result.

2. The method according to claim 1, characterized in that Acquiring the second perception data includes: Acquire third perception data of the target obstacle collected by the laser radar at the first collection moment; Point cloud compensation is performed on the third perception data based on the first acquisition time, the second acquisition time and the point cloud acquisition frequency to obtain the second perception data, wherein the point cloud acquisition frequency is the acquisition frequency of the third perception data acquired by the laser radar, the second acquisition time is the time when the camera acquires the first perception data, and the second perception data is aligned with the perception data of the first perception data at the second acquisition time.

3. The method according to claim 1, characterized in that The feature fusion of the semantic feature and the structural feature based on the target correspondence relationship to obtain the fused feature includes: Determine a relative position of first position information and second position information, wherein the first position information is the installation position information of the camera, and the second position information is the installation position information of the laser radar; Projecting the second perception data onto a normalized plane based on the relative position to obtain the target correspondence relationship; The semantic feature and the structural feature are fused based on a preset fusion algorithm and the target correspondence relationship to obtain the fused feature.

4. The method according to claim 1, characterized in that The method further comprises: Planning an obstacle avoidance path for the target vehicle based on the detection result, a preset planning algorithm, the driving scenario and driving requirements of the target vehicle; The target vehicle is controlled to travel based on the obstacle avoidance path.

5. The method according to any one of claims 1 to 4, characterized in that Extracting the semantic features of the target obstacle from the first perception data includes: Based on a preset image segmentation algorithm, feature extraction is performed on the first perception data to obtain the semantic feature, wherein the semantic feature includes an obstacle type feature and an obstacle color feature.

6. The method according to any one of claims 1 to 4, characterized in that Extracting the structural features of the target obstacle from the second perception data includes: Based on a preset point cloud segmentation algorithm, feature extraction is performed on the second perception data to obtain the structural feature, wherein the structural feature includes obstacle position feature and obstacle shape feature.

7. The method according to any one of claims 1 to 4, characterized in that Before extracting the semantic features of the target obstacle from the first perception data and extracting the structural features of the target obstacle from the second perception data, the method further includes: Data denoising and data enhancement are performed on the first perception data and the second perception data.

8. The method according to any one of claims 1 to 4, characterized in that The method further comprises: The detection result is displayed on a display of the target vehicle.

9. The method according to any one of claims 1 to 4, characterized in that: The cameras are installed on the front, rear, left and right sides of the target vehicle, and the laser radar is installed on the top of the target vehicle.

10. An obstacle detection device, characterized in that: include: An acquisition module, used to acquire first perception data and second perception data of a target vehicle, wherein the first perception data is image data of a target obstacle acquired by a camera, the second perception data is point cloud data of the target obstacle acquired by a laser radar, and the target obstacle is an obstacle within a preset range of the target vehicle; an extraction module, configured to extract semantic features of the target obstacle from the first perception data, and extract structural features of the target obstacle from the second perception data; a fusion module, configured to perform feature fusion on the semantic feature and the structural feature based on a target correspondence relationship to obtain a fusion feature, wherein the target correspondence relationship is an association relationship between the first perception data and the second perception data; The detection module is used to detect the target obstacle based on the fusion feature to obtain a detection result.

11. A vehicle, characterized in that: The vehicle is used to execute the obstacle detection method described in any one of claims 1 to 9 above.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the obstacle detection method described in any one of claims 1 to 9 when running on a computer or a processor.

13. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to run the computer program to execute the obstacle detection method described in any one of claims 1 to 9.

14. A computer program product, characterized in that The invention comprises a computer program, which implements the obstacle detection method described in any one of claims 1 to 9 when being executed by a processor.

Images