Target detection method, device, vehicle and storage medium

Abstract

The present disclosure relates to a target detection method, apparatus, vehicle, and storage medium. The method comprises: acquiring environmental perception data from a vehicle, the environmental perception data being data obtained by detecting the environment surrounding the vehicle; and processing the environmental perception data using a target detection model to obtain target detection results, the target detection results including a road segmentation result, a first obstacle detection result corresponding to a driving scenario, and a second obstacle detection result corresponding to a parking scenario. By processing the vehicle's environmental perception data using a target detection model trained with sample environmental perception data carrying multiple labels, the vehicle can obtain a road segmentation result, a first obstacle detection result corresponding to a driving scenario, and a second obstacle detection result corresponding to a parking scenario. This avoids switching models between different driving scenarios and the impact of model switching on autonomous driving.

Description

Technical Field

[0001] The present disclosure relates to the field of autonomous driving technology, and in particular to a target detection method, device, vehicle, and storage medium. Background Art

[0002] In autonomous driving scenarios, different target detection and road segmentation requirements exist for different vehicle driving scenarios, so as to obtain detection results corresponding to the vehicle driving scenarios and determine the corresponding autonomous driving strategies.

[0003] In related technologies, detection results of corresponding vehicle driving scenarios are obtained through multiple independent models. During this process, changes in vehicle driving scenarios require switching corresponding models. In the process of switching models, the vehicle's automatic driving function will be suspended, affecting the vehicle's automatic driving effect. Summary of the Invention

[0004] To overcome the problems existing in the related art, the present disclosure provides a target detection method, device, vehicle and storage medium. By processing the vehicle's environmental perception data with a target detection model obtained by training sample environmental perception data carrying multiple labels, a road segmentation result, a first obstacle detection result corresponding to the driving scene and a second obstacle detection result corresponding to the parking scene can be obtained, thereby avoiding the switching of models in different driving scenarios and avoiding the impact of switching models on autonomous driving.

[0005] According to a first aspect of an embodiment of the present disclosure, there is provided a target detection method, comprising:

[0006] Acquiring environmental perception data of the vehicle, wherein the environmental perception data is data obtained by detecting the environment around the vehicle;

[0007] The environmental perception data is processed by a target detection model to obtain a target detection result, which includes a road segmentation result, a first obstacle detection result corresponding to a driving scene, and a second obstacle detection result corresponding to a parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data, and each sample environmental perception data carries multiple labels.

[0008] Optionally, the target detection model includes a shared recognition network, a road segmentation determination network, a first obstacle determination network, and a second obstacle determination network. Processing the environment perception data by the target detection model to obtain a target detection result includes:

[0009] Processing the environmental perception data through the shared recognition network to obtain a global feature vector;

[0010] Processing the global feature vector using the first obstacle determination network to obtain the first obstacle detection result;

[0011] Processing the global feature vector using the second obstacle determination network to obtain a second obstacle detection result;

[0012] The environmental perception data is processed by a road segmentation determination network to obtain the road segmentation result.

[0013] Optionally, the shared recognition network includes a first feature extraction network and a second feature extraction network;

[0014] Before processing the environmental perception data through the shared recognition network to obtain a global feature vector, the method further includes:

[0015] Performing voxel processing on the environmental perception data to obtain voxelized data;

[0016] The processing of the environmental perception data by the shared recognition network to obtain a global feature vector includes:

[0017] Performing preliminary feature extraction on the voxelized data through the first feature extraction network to obtain a multi-layer feature vector;

[0018] The second feature extraction network performs deep feature extraction on the multi-layer feature vector to obtain the global feature vector.

[0019] Optionally, the first obstacle determination network includes a vehicle head network and a vehicle target output network;

[0020] The processing of the global feature vector by the first obstacle determination network to obtain the first obstacle detection result includes:

[0021] Processing the global feature vector through the vehicle head network to obtain a vehicle target vector corresponding to the vehicle detection target;

[0022] The driving target vector is processed by the driving target output network to obtain the first obstacle detection result.

[0023] Optionally, the second obstacle determination network includes a parking head network and a parking target output network;

[0024] The processing of the global feature vector by the second obstacle determination network to obtain the second obstacle detection result includes:

[0025] Processing the global feature vector through the parking head network to obtain a parking target vector corresponding to the parking detection target;

[0026] The parking target vector is processed by the parking target output network to obtain the second obstacle detection result.

[0027] Optionally, the road segmentation determination network includes a fully connected neural network and a road output network;

[0028] The environmental perception data is processed by a road segmentation determination network to obtain the road segmentation result, including:

[0029] Processing the environmental perception data through the fully connected neural network to obtain a road surface feature vector;

[0030] The road surface feature vector is processed by the road surface output network to obtain the road surface segmentation result.

[0031] Optionally, the fully connected neural network is connected to the driving head network;

[0032] The processing of the environmental perception data by the fully connected neural network to obtain a road surface feature vector includes:

[0033] The environmental perception data and the driving target vector are processed by the fully connected neural network to obtain a road surface feature vector.

[0034] Optionally, the method further includes:

[0035] Acquiring a driving scene of a vehicle, wherein the driving scene includes a driving scene and a parking scene;

[0036] determining a target driving strategy according to the driving scenario and the target detection result;

[0037] The vehicle is controlled according to the target driving strategy.

[0038] According to a second aspect of an embodiment of the present disclosure, there is provided an object detection device, including:

[0039] A first acquisition module is configured to acquire environmental perception data of the vehicle, where the environmental perception data is data obtained by detecting the environment around the vehicle;

[0040] The first acquisition module is configured to process the environmental perception data through a target detection model to obtain a target detection result, wherein the target detection result includes a road segmentation result, a first obstacle detection result corresponding to a driving scene, and a second obstacle detection result corresponding to a parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data, and each sample environmental perception data carries multiple labels.

[0041] According to a third aspect of an embodiment of the present disclosure, there is provided a vehicle, comprising:

[0042] processor;

[0043] a memory for storing processor-executable instructions;

[0044] The processor is configured to implement the steps of the target detection method provided in the first aspect of the present disclosure.

[0045] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the target detection method provided in the first aspect of the present disclosure are implemented.

[0046] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0047] By acquiring the vehicle's environmental perception data, which is data obtained by detecting the environment around the vehicle, and processing the environmental perception data through a target detection model, a target detection result is obtained, which includes a road segmentation result, a first obstacle detection result corresponding to the driving scene, and a second obstacle detection result corresponding to the parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data. Each sample environmental perception data carries multiple labels. By processing the vehicle's environmental perception data with a target detection model obtained by training the sample environmental perception data carrying multiple labels, a road segmentation result, a first obstacle detection result corresponding to the driving scene, and a second obstacle detection result corresponding to the parking scene can be obtained, thereby avoiding switching of models under different driving scenarios and avoiding the impact of switching models on autonomous driving.

[0048] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0050] Figure 1 The figure is a flowchart of a target detection method according to an exemplary embodiment.

[0051] Figure 2 The figure is a flowchart of a method for obtaining a target detection result according to an exemplary embodiment.

[0052] Figure 3 The figure is a flowchart of a method for obtaining a global feature vector according to an exemplary embodiment.

[0053] Figure 4 The figure is a flowchart of a method for obtaining a first obstacle detection result according to an exemplary embodiment.

[0054] Figure 5 The figure is a flowchart of a method for obtaining a second obstacle detection result according to an exemplary embodiment.

[0055] Figure 6 The figure is a schematic structural diagram of a target detection model according to an exemplary embodiment.

[0056] Figure 7 The figure is a block diagram of a target detection device according to an exemplary embodiment.

[0057] Figure 8 is a block diagram of a vehicle according to an exemplary embodiment. DETAILED DESCRIPTION

[0058] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0059] Figure 1 FIG. 1 is a flow chart of a target detection method according to an exemplary embodiment. Figure 1 As shown, the following steps are included.

[0060] In step S101 , environmental perception data of the vehicle is acquired, where the environmental perception data is data obtained by detecting the environment around the vehicle.

[0061] In this embodiment, the vehicle's surrounding environment can be detected by an onboard sensor to obtain the vehicle's environmental perception data. The onboard sensor can be a laser radar, and the environmental perception data can be point cloud data.

[0062] In step S102, the environmental perception data is processed by the target detection model to obtain a target detection result, which includes a road segmentation result, a first obstacle detection result corresponding to the driving scene, and a second obstacle detection result corresponding to the parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data, and each sample environmental perception data carries multiple labels.

[0063] In this embodiment, the basic network can be iteratively trained multiple times using multiple sample environmental perception data to obtain a trained object detection model. Each sample environmental perception data carries multiple labels. For example, the labels may include the annotated true road segmentation result corresponding to the sample environmental perception data, the first true obstacle result corresponding to the driving scene, and the second true obstacle result corresponding to the parking scene. This allows the trained object detection model to take environmental perception data as input and output a road segmentation result corresponding to the environmental perception data, a first obstacle detection result corresponding to the driving scene, and a second obstacle detection result corresponding to the parking scene. This object detection model is a multi-task model. After training with multiple sample environmental perception data carrying multiple labels, it can achieve stronger feature extraction capabilities, thereby improving the accuracy of the output object detection results.

[0064] The disclosed embodiment obtains environmental perception data from the vehicle, which is data obtained by detecting the environment around the vehicle, and processes the environmental perception data through a target detection model to obtain a target detection result. The target detection result includes a road segmentation result, a first obstacle detection result corresponding to the driving scene, and a second obstacle detection result corresponding to the parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data. Each sample environmental perception data carries multiple labels. The target detection model trained with the sample environmental perception data carrying multiple labels is used to process the vehicle's environmental perception data to obtain a road segmentation result, a first obstacle detection result corresponding to the driving scene, and a second obstacle detection result corresponding to the parking scene. This avoids switching models under different driving scenarios and avoids the impact of switching models on autonomous driving. Moreover, the multi-task target detection model of the disclosed embodiment can save storage resources and computing resources.

[0065] Figure 2 FIG. 1 is a flow chart showing a method for obtaining target detection results according to an exemplary embodiment. Figure 2As shown, in one possible implementation, the target detection model may include a shared recognition network, a road segmentation determination network, a first obstacle determination network, and a second obstacle determination network. Processing the environment perception data through the target detection model to obtain a target detection result may include the following steps:

[0066] In step S201, the environmental perception data is processed through a shared recognition network to obtain a global feature vector.

[0067] In this embodiment, the shared recognition network is used to perform feature extraction on the environmental perception data, so as to extract the features of each object in the environmental perception data and obtain a global feature vector.

[0068] In step S202, the global feature vector is processed by a first obstacle determination network to obtain a first obstacle detection result.

[0069] In this embodiment, the first obstacle determination network can identify and further extract features from the global feature vector to obtain a feature vector related to the driving scenario from the global feature vector, and then regress this feature vector to obtain a first obstacle detection result corresponding to the driving scenario. The first obstacle detection result can be the detection results of multiple first target obstacles within a first preset distance. The first target obstacles can be cars, trucks, bicycles, pedestrians, etc. The first preset distance can be 40 meters to the left, 40 meters to the right, and 150 meters in front of the vehicle. The first obstacle detection result can include the position, speed, and size of the multiple first target obstacles within the first preset distance.

[0070] In step S203, the global feature vector is processed by the second obstacle determination network to obtain a second obstacle detection result.

[0071] In this embodiment, the second obstacle determination network can identify and further extract features from the global feature vector to derive a feature vector related to the parking scenario from the global feature vector. This feature vector is then regressed to produce a second obstacle detection result corresponding to the parking scenario. The second obstacle detection result can include detection results for multiple second target obstacles within a second preset distance. These second target obstacles can include cars, pillars, pedestrians, and so on. The second preset distance can be 20 meters to the left, right, front, and rear of the vehicle. The second obstacle detection result can include the position, velocity, and size of the multiple second target obstacles within the second preset distance.

[0072] In step S204, the environmental perception data is processed by the road segmentation determination network to obtain a road segmentation result.

[0073] In this embodiment, the road segmentation determination network can extract features from environmental perception data and identify road features, so as to output a road segmentation result, which can be a vehicle driving area where the vehicle can drive.

[0074] By coordinating a shared recognition network with a road segmentation determination network, a first obstacle determination network, and a second obstacle determination network corresponding to multiple tasks, a target detection network is obtained, so that one model can complete multiple tasks. After being trained with multiple sample environmental perception data carrying multiple labels, the shared recognition network can obtain stronger feature extraction capabilities, thereby improving the accuracy of the output global feature vector and saving storage resources and computing resources.

[0075] In a possible implementation, before processing the environmental perception data through a shared recognition network to obtain a global feature vector, the environmental perception data may also be preprocessed. For example, if the environmental perception data is point cloud data, the point cloud data may be voxelized to obtain voxelized data, thereby making the disordered point cloud data ordered to facilitate subsequent feature extraction.

[0076] Figure 3 is a flow chart showing a method for obtaining a global feature vector according to an exemplary embodiment. Figure 3 As shown, in a possible implementation, the shared recognition network includes a first feature extraction network and a second feature extraction network. Processing the environmental perception data through the shared recognition network to obtain a global feature vector may include the following steps:

[0077] In step S301, a first feature extraction network is used to perform preliminary feature extraction on the voxelized data to obtain a multi-layer feature vector.

[0078] In this embodiment, the first feature extraction network may be a backbone network, such as a CNN (Convolutional Neural Network), which may perform preliminary feature extraction on the voxelized data to obtain multi-layer feature vectors, which are feature vectors of different dimensions.

[0079] In step S302, deep feature extraction is performed on the multi-layer feature vectors through a second feature extraction network to obtain a global feature vector.

[0080] In this embodiment, the second feature extraction network can be FPN (Feature Pyramid Network), which can perform deep feature extraction on multi-layer feature vectors through FPN, and obtain a global feature vector by combining the feature information of each layer in the multi-layer feature vectors for processing, so as to facilitate subsequent target detection.

[0081] Figure 4 FIG. 1 is a flow chart showing a method for obtaining a first obstacle detection result according to an exemplary embodiment. Figure 4 As shown, in one possible implementation, the first obstacle determination network includes a vehicle head network and a vehicle target output network; processing the global feature vector by the first obstacle determination network to obtain a first obstacle detection result may include the following steps:

[0082] In step S401, the global feature vector is processed by the vehicle head network to obtain a vehicle target vector corresponding to the vehicle detection target.

[0083] In this embodiment, the global feature vector can be screened and feature extracted through the driving head network to obtain a driving target vector, wherein the driving detection target is the first target obstacle, which can be a car, truck, bicycle, pedestrian, etc.

[0084] In step S402, the driving target vector is processed by the driving target output network to obtain a first obstacle detection result.

[0085] In this embodiment, the driving target vector is regressed through the driving target output network to obtain the first obstacle detection result, which may be the position, speed and size of multiple first target obstacles within the first preset distance.

[0086] Figure 5 FIG. 1 is a flow chart showing a method for obtaining a second obstacle detection result according to an exemplary embodiment. Figure 5 As shown, in one possible implementation, the second obstacle determination network includes a parking head network and a parking target output network. The second obstacle determination network processes the global feature vector to obtain a second obstacle detection result, including:

[0087] In step S501 , the global feature vector is processed by the parking head network to obtain a parking target vector corresponding to the parking detection target.

[0088] In this embodiment, the global feature vector can be screened and feature extracted through the parking head network to obtain a parking target vector, wherein the parking detection target is the second target obstacle, which can be a car, a pillar, a pedestrian, etc.

[0089] In step S502 , the parking target vector is processed by the parking target output network to obtain a second obstacle detection result.

[0090] In this embodiment, the parking target vector is regressed by the parking target output network to obtain a second obstacle detection result. The second obstacle detection result may be the position, speed, and size of multiple second target obstacles within the second preset distance.

[0091] In one possible implementation, the road surface segmentation determination network includes a fully connected neural network and a road surface output network; processing the environmental perception data through the road surface segmentation determination network to obtain a road surface segmentation result may include: processing the environmental perception data through the fully connected neural network to obtain a road surface feature vector; processing the road surface feature vector through the road surface output network to obtain a road surface segmentation result.

[0092] In this embodiment, a fully connected neural network can be used to extract features from environmental perception data and identify road features so that a road feature vector can be output. The road feature vector is then regressed through a road output network to obtain a road segmentation result, which can be a vehicle driving area where the vehicle can travel.

[0093] Figure 6 FIG. 1 is a schematic diagram showing a structure of a target detection model according to an exemplary embodiment. Figure 6 As shown, in one possible implementation, a fully connected neural network is connected to a driving head network; the environmental perception data is processed by the fully connected neural network to obtain a road surface feature vector, including: processing the environmental perception data and the driving target vector by the fully connected neural network to obtain a road surface feature vector.

[0094] In this embodiment, the fully connected neural network is connected to the driving head network, and the driving head network can transmit the obtained driving target vector to the fully connected neural network so that the fully connected neural network can fuse the environmental perception data and the driving target vector. By fusing the high-resolution driving target vector and the low-resolution environmental network perception data, the semantic information input into the fully connected neural network can be enriched, thereby obtaining a more accurate road feature vector.

[0095] In one possible implementation, after obtaining the target detection result, the vehicle's driving scene can be acquired, including driving scenes and parking scenes; a target driving strategy is determined based on the driving scene and the target detection result; and the vehicle is controlled according to the target driving strategy.

[0096] In this embodiment, the vehicle's current driving scenario is obtained. The current driving scenario can be either a driving scenario or a parking scenario. If the driving scenario is a driving scenario, a target driving strategy (which can include the vehicle's driving speed and path planning) is determined based on the road segmentation result in the target detection results and the first obstacle detection result corresponding to the driving scenario. The vehicle is then controlled to perform autonomous driving based on the obtained target driving strategy. If the driving scenario is a parking scenario, the target driving strategy is determined based on the road segmentation result in the target detection results and the second obstacle detection result corresponding to the parking scenario.

[0097] Figure 7 FIG. 1 is a block diagram of a target detection device according to an exemplary embodiment. Figure 7 The target detection device 700 includes a first acquisition module 701 and a first obtaining module 702.

[0098] The first acquisition module 701 is configured to acquire environmental perception data of the vehicle, where the environmental perception data is data obtained by detecting the environment around the vehicle;

[0099] The first acquisition module 702 is configured to process the environmental perception data through a target detection model to obtain a target detection result, wherein the target detection result includes a road segmentation result, a first obstacle detection result corresponding to a driving scene, and a second obstacle detection result corresponding to a parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data, and each sample environmental perception data carries multiple labels.

[0100] Optionally, the target detection model includes a shared recognition network, a road segmentation determination network, a first obstacle determination network, and a second obstacle determination network, and the first obtaining module 702 includes:

[0101] A first acquisition submodule is configured to process the environmental perception data through the shared recognition network to obtain a global feature vector;

[0102] a second obtaining submodule, configured to process the global feature vector through the first obstacle determination network to obtain the first obstacle detection result;

[0103] a third obtaining submodule, configured to process the global feature vector through the second obstacle determination network to obtain the second obstacle detection result;

[0104] The fourth obtaining submodule is configured to process the environmental perception data through a road surface segmentation determination network to obtain the road surface segmentation result.

[0105] Optionally, the shared recognition network includes a first feature extraction network and a second feature extraction network;

[0106] The target detection device 700 further includes:

[0107] A second obtaining module is configured to perform voxelization processing on the environmental perception data to obtain voxelized data;

[0108] The first obtaining submodule includes:

[0109] a first obtaining unit, configured to perform preliminary feature extraction on the voxelized data through the first feature extraction network to obtain a multi-layer feature vector;

[0110] The second obtaining unit is configured to perform deep feature extraction on the multi-layer feature vector through the second feature extraction network to obtain the global feature vector.

[0111] Optionally, the first obstacle determination network includes a vehicle head network and a vehicle target output network;

[0112] The second obtaining submodule includes:

[0113] A third obtaining unit is configured to process the global feature vector through the vehicle head network to obtain a vehicle target vector corresponding to the vehicle detection target;

[0114] The fourth obtaining unit is configured to process the driving target vector through the driving target output network to obtain the first obstacle detection result.

[0115] Optionally, the second obstacle determination network includes a parking head network and a parking target output network;

[0116] The third obtaining submodule includes:

[0117] a fifth obtaining unit configured to process the global feature vector through the parking head network to obtain a parking target vector corresponding to the parking detection target;

[0118] A sixth obtaining unit is configured to process the parking target vector through the parking target output network to obtain the second obstacle detection result.

[0119] Optionally, the road segmentation determination network includes a fully connected neural network and a road output network;

[0120] The fourth obtaining submodule includes:

[0121] a seventh obtaining unit, configured to process the environmental perception data through the fully connected neural network to obtain a road surface feature vector;

[0122] An eighth obtaining unit is configured to process the road surface feature vector through the road surface output network to obtain the road surface segmentation result.

[0123] Optionally, the fully connected neural network is connected to the driving head network;

[0124] The seventh obtaining unit includes:

[0125] The acquisition subunit is configured to process the environmental perception data and the driving target vector through the fully connected neural network to obtain the road surface feature vector.

[0126] Optionally, the target detection device 700 further includes:

[0127] A second acquisition module is configured to acquire a driving scene of the vehicle, wherein the driving scene includes a driving scene and a parking scene;

[0128] a determination module, configured to determine a target driving strategy based on the driving scenario and the target detection result;

[0129] A control module is configured to control the vehicle according to the target driving strategy.

[0130] Regarding the target detection device 700 in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0131] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, which implement the steps of the target detection method provided by the present disclosure when the program instructions are executed by a processor.

[0132] The present disclosure also provides a vehicle, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the steps of the target detection method described in the above embodiment.

[0133] Figure 8 8 is a block diagram illustrating a vehicle according to an exemplary embodiment. For example, vehicle 800 may be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or another type of vehicle. Vehicle 800 may be an autonomous vehicle or a semi-autonomous vehicle.

[0134] Reference Figure 8 Vehicle 800 may include various subsystems, such as an infotainment system 810, a perception system 820, a decision-making and control system 830, a drive system 840, and a computing platform 850. Vehicle 800 may also include more or fewer subsystems, and each subsystem may include multiple components. Furthermore, each subsystem and each component of vehicle 800 may be interconnected via wired or wireless means.

[0135] In some embodiments, the infotainment system 810 may include a communication system, an entertainment system, a navigation system, and the like.

[0136] The perception system 820 may include several sensors for sensing information about the environment surrounding the vehicle 800. For example, the perception system 820 may include a global positioning system (which may be a GPS system, a BeiDou system, or another positioning system), an inertial measurement unit (IMU), a laser radar, a millimeter-wave radar, an ultrasonic radar, and a camera.

[0137] The decision control system 830 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0138] The drive system 840 may include components that provide power to the vehicle 800. In one embodiment, the drive system 840 may include an engine, a power source, a transmission system, and wheels. The engine may be an internal combustion engine, an electric motor, an air compression engine, or a combination thereof. The engine is capable of converting energy provided by the power source into mechanical energy.

[0139] Some or all functions of the vehicle 800 are controlled by a computing platform 850. The computing platform 850 may include at least one processor 851 and a memory 852. The processor 851 may execute instructions 853 stored in the memory 852.

[0140] The processor 851 can be any conventional processor, such as a commercially available CPU. The processor can also include a graphics processor (GPU), a field programmable gate array (FPGA), a system on chip (SOC), an application specific integrated circuit (ASIC), or a combination thereof.

[0141] The memory 852 can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0142] In addition to instructions 853 , memory 852 may also store data, such as road maps, route information, and vehicle location, direction, speed, etc. The data stored in memory 852 may be used by computing platform 850 .

[0143] In the embodiment of the present disclosure, the processor 851 may execute the instruction 853 to complete all or part of the steps of the above-mentioned target detection method.

[0144] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program executable by a programmable device, and has a code portion for performing the above-mentioned target detection method when executed by the programmable device.

[0145] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the present disclosure. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0146] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A target detection method, characterized in that: include: Acquire environmental perception data of the vehicle, wherein the environmental perception data is point cloud data obtained by detecting the environment around the vehicle through a laser radar; The environmental perception data is processed by a target detection model to obtain a target detection result, which includes a road segmentation result, a first obstacle detection result corresponding to a driving scene, and a second obstacle detection result corresponding to a parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data. Each sample environmental perception data carries multiple labels, and the labels are the true road segmentation result corresponding to the sample environmental perception data, the true first obstacle result corresponding to the driving scene, and the true second obstacle result corresponding to the parking scene.

2. The target detection method according to claim 1, wherein: The target detection model includes a shared recognition network, a road segmentation determination network, a first obstacle determination network, and a second obstacle determination network. The target detection model processes the environmental perception data to obtain a target detection result, including: Processing the environmental perception data through the shared recognition network to obtain a global feature vector; Processing the global feature vector using the first obstacle determination network to obtain the first obstacle detection result; Processing the global feature vector using the second obstacle determination network to obtain a second obstacle detection result; The environmental perception data is processed by the road surface segmentation determination network to obtain the road surface segmentation result.

3. The target detection method according to claim 2, characterized in that: The shared recognition network includes a first feature extraction network and a second feature extraction network; Before processing the environmental perception data through the shared recognition network to obtain a global feature vector, the method further includes: Performing voxel processing on the environmental perception data to obtain voxelized data; The processing of the environmental perception data by the shared recognition network to obtain a global feature vector includes: Performing preliminary feature extraction on the voxelized data through the first feature extraction network to obtain a multi-layer feature vector; The second feature extraction network performs deep feature extraction on the multi-layer feature vector to obtain the global feature vector.

4. The target detection method according to claim 2, wherein: The first obstacle determination network includes a vehicle head network and a vehicle target output network; The processing of the global feature vector by the first obstacle determination network to obtain the first obstacle detection result includes: Processing the global feature vector through the vehicle head network to obtain a vehicle target vector corresponding to the vehicle detection target; The driving target vector is processed by the driving target output network to obtain the first obstacle detection result.

5. The target detection method according to claim 2, wherein: The second obstacle determination network includes a parking head network and a parking target output network; The processing of the global feature vector by the second obstacle determination network to obtain the second obstacle detection result includes: Processing the global feature vector through the parking head network to obtain a parking target vector corresponding to the parking detection target; The parking target vector is processed by the parking target output network to obtain the second obstacle detection result.

6. The target detection method according to claim 4, characterized in that: The road segmentation determination network includes a fully connected neural network and a road output network; The environmental perception data is processed by a road segmentation determination network to obtain the road segmentation result, including: Processing the environmental perception data through the fully connected neural network to obtain a road surface feature vector; The road surface feature vector is processed by the road surface output network to obtain the road surface segmentation result.

7. The target detection method according to claim 6, characterized in that: The fully connected neural network is connected to the driving head network; The processing of the environmental perception data by the fully connected neural network to obtain a road surface feature vector includes: The environmental perception data and the driving target vector are processed by the fully connected neural network to obtain a road surface feature vector.

8. The target detection method according to claim 1, wherein: The method further comprises: Acquiring a driving scene of a vehicle, wherein the driving scene includes a driving scene and a parking scene; determining a target driving strategy according to the driving scenario and the target detection result; The vehicle is controlled according to the target driving strategy.

9. A target detection device, characterized in that: include: A first acquisition module is configured to acquire environmental perception data of the vehicle, wherein the environmental perception data is point cloud data obtained by detecting the environment around the vehicle through a laser radar; The first acquisition module is configured to process the environmental perception data through a target detection model to obtain a target detection result, wherein the target detection result includes a road segmentation result, a first obstacle detection result corresponding to a driving scene, and a second obstacle detection result corresponding to a parking scene. The target detection model is obtained by training a basic network based on multiple sample environmental perception data. Each sample environmental perception data carries multiple labels, and the labels are the true road segmentation result corresponding to the sample environmental perception data, the first obstacle true result corresponding to the driving scene, and the second obstacle true result corresponding to the parking scene.

10. A vehicle, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to execute the steps of the target detection method according to any one of claims 1 to 8.

11. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by a processor, the steps of the target detection method described in any one of claims 1 to 8 are implemented.

Images