Traffic target detection method

Abstract

The invention provides a traffic target detection method. The method is applied to the technical field of traffic target detection and comprises the steps of collecting radar data and video data of a target vehicle and traffic facilities; performing accurate time alignment on the video and the radar data to ensure that the timestamps of the data are consistent; image video data and radar data are subjected to preliminary matching, visual data are used for target segmentation, a segmentation result and the radar data are subjected to target matching, and low-level fusion of a target vehicle is completed; performing target feature extraction and fusion by using a convolutional neural network in combination with motion information in the radar data and image information in the visual data to realize semantic-level fusion; and real-time processing and feedback: inputting the fused data into the NPU to execute real-time target detection and behavior analysis to obtain a target type, a motion track and a behavior track, and providing traffic guidance and prediction for a target vehicle. In this way, the accuracy of traffic target detection can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic target detection, and in particular to a traffic target detection method. Background Art

[0002] Traffic target detection aims to automatically identify and locate objects in traffic scenes, and provide support for traffic safety, road management, and autonomous driving. Traffic target detection technology can automatically and in real time identify various objects in traffic scenes, including pedestrians, vehicles, traffic signs, and traffic lights, and accurately track their location, speed, and other information, which can help traffic management systems quickly identify potential safety hazards, warn of dangerous situations, and effectively reduce the occurrence of traffic accidents; through real-time detection and analysis of traffic flow, traffic management departments can dynamically adjust traffic lights and optimize traffic signal cycles according to changes in vehicle and pedestrian flow, thereby improving road capacity and reducing traffic congestion; traffic target detection can not only identify vehicles and pedestrians, but also combine environmental monitoring sensors to conduct real-time detection of air quality, noise pollution, etc., provide comprehensive traffic environment data, and help governments and relevant departments to keep abreast of changes in urban traffic environment and take corresponding measures to reduce pollution and adverse effects caused by traffic.

[0003] At present, traditional traffic target detection methods rely on a single vehicle-mounted radar or camera sensor. A single visual sensor cannot accurately obtain target information, and a single radar sensor relies on hypothetical assumptions about the environment and cannot effectively deal with scenes that the sensor cannot observe. Vehicle-mounted target detection technology based on vision and radar fusion is a detection method that has developed rapidly in recent years. It aims to use multiple sensors for feature extraction and fusion, and to achieve accurate target detection by exploring the complementary advantages of multiple sensors. However, traditional fusion methods have problems such as insufficient target segmentation, low accuracy, and high latency. Summary of the invention

[0004] The invention provides a traffic target detection method.

[0005] According to a first aspect of the present invention, a method for detecting a traffic target is provided. The method comprises:

[0006] Use vehicle radar sensors and video sensors installed on edge nodes to collect radar data and video data of target vehicles and traffic facilities;

[0007] Accurately time-align video and radar data to ensure consistent timestamps of the data;

[0008] Use image video data and radar data for preliminary matching, use visual data for target segmentation, match the segmented results with radar data, and complete the low-level fusion of the target vehicle;

[0009] Using convolutional neural networks, the motion information in radar data and the image information in visual data are combined to extract and fuse target features, thus achieving semantic-level fusion.

[0010] Real-time processing and feedback The above fused data is input into the NPU to perform real-time target detection and behavior analysis, obtain the target type, motion trajectory and behavior trajectory, and provide traffic guidance and prediction for the target vehicle.

[0011] Furthermore: the radar data includes vehicle speed, acceleration, heading angle, length and width; the video data includes shape, color and texture; the radar data and video data are preprocessed through edge computing, including: denoising, filtering, signal enhancement of radar data and denoising, enhancement and feature extraction of image data.

[0012] Furthermore, the edge nodes of the MEC platform are used to deploy NTP servers for time synchronization, so that the collection, processing and fusion of all video and radar data in the system are based on the same clock; - The MEC platform is used to perform spatial alignment of image frames or point clouds of radar data and video data to form a spatial coordinate mapping matrix, and complete the alignment of the same target in different spatial dimensions of the two data.

[0013] Furthermore: Use YOLO, SSD or Faster RCNN deep learning target detection models to perform preliminary target detection on the spatial position of the target in the radar data and the image in the visual data to form the target category and regional position of the image; based on the target position and category obtained in the image target detection, use the edge nodes of the MEC platform to perform target matching, find the corresponding target in the radar data, and form preliminary matching data.

[0014] Furthermore: based on the preliminary matching data, the video target category and the physical information in the radar data are used to perform data fusion, extract target features and predict target behavior; the motion information in the radar data, the image information in the visual data, and the information for predicting target behavior are combined for high-level fusion to obtain the target's behavior trajectory and motion trajectory for traffic monitoring.

[0015] Furthermore: high-level semantic fusion is completed at the MEC processing edge node, and the fused data is sent to the NPU for real-time target detection and behavior analysis; the NPU is used to perform real-time target detection and behavior analysis, and output the target type, motion trajectory and behavior trajectory; the above data is fed back to the third-party receiving platform for traffic guidance and prediction.

[0016] Furthermore: Leverage the MEC platform in conjunction with NPU hardware to accelerate the execution of deep learning target detection and classification algorithms.

[0017] According to a second aspect of the present invention, an electronic device is provided, comprising: a memory and a processor, wherein a computer program is stored in the memory, and the method is implemented when the processor executes the program.

[0018] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method when executed by a processor.

[0019] Beneficial effects of the present invention:

[0020] 1. The present invention combines the target detection algorithm based on deep learning with the MEC platform, uses the MEC platform to accelerate the target detection algorithm, and realizes the rapid output of target detection results, thereby greatly improving the real-time and accuracy of target detection.

[0021] 2. The present invention combines the target detection results with the radar data for preliminary fusion, uses the radar data as verification of the target authenticity, combines the image with the convolutional neural network for feature extraction, combines the radar data such as position and angle with the category and features as parameter information, and performs preliminary fusion of the vehicle trajectory, thereby achieving accurate prediction of vehicle behavior and providing a reference for traffic management.

[0022] 3. The present invention combines traffic edge computing and deploys the target perception and detection algorithms on the MEC platform. It does not require a large amount of computing resources and uses NPU hardware acceleration to achieve real-time target detection. The version of the deployment algorithm can be quickly upgraded through the edge computing platform, thereby improving the deployment efficiency of the algorithm.

[0023] 4. The present invention combines the technical characteristics of 5G MEC, deploys the target detection algorithm at the MEC edge node, and quickly completes target detection through target data collection and fusion. It is especially suitable for detection scenarios with strong real-time performance.

[0024] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and other features, advantages and aspects of the embodiments of the present invention will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are used to better understand the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0026] Figure 1 A flow chart of a traffic target detection method according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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 are within the scope of protection of the present invention.

[0028] In addition, the term "and / or" in this article is only a description of the association relationship between the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0029] Figure 1 The flow chart of the traffic target detection method according to an embodiment of the present invention is shown. The method includes: S101, using the vehicle radar sensor and video sensor installed on the edge node to collect radar data and video data of the target vehicle and traffic facilities; S102, accurately aligning the video and radar data in time to ensure the consistency of the timestamp of the data; S103, using image video data and radar data for preliminary matching, using visual data for target segmentation, matching the segmented results with radar data for target matching, and completing the low-level fusion of the target vehicle; S104, using convolutional neural network, combining the motion information in the radar data with the image information in the visual data to extract and fuse the target features, and realize semantic-level fusion; S105, real-time processing and feedback, inputting the above fusion data into NPU to perform real-time target detection and behavior analysis, obtain the target type, motion trajectory and behavior trajectory, and provide traffic guidance and prediction for the target vehicle.

[0030] The present invention provides a traffic target detection method, which has the advantage of being able to accurately and quickly provide feedback on vehicle behavior on the road, and is of great significance for real-time vehicle detection and behavior analysis.

[0031] The present invention is described based on a system of 5G edge computing, video image processing and deep learning: target detection algorithms based on deep learning usually require the collection of a large amount of training data and a complex training process, which requires a high amount of calculation, seriously reducing the real-time performance of data processing, which is particularly disadvantageous for applications with strong real-time requirements. At the same time, video image target detection is usually processed based on images and videos. Due to the influence of target motion uncertainty, the target visual detection results are unstable. In the hardware deployment scheme based on edge computing target detection, only edge node computing or edge computing and cloud center computing are often considered, and the computing power of edge computing and cloud computing cannot be taken into account at the same time. Therefore, the present invention combines 5G MEC technology and proposes a target detection method and device suitable for MEC scenarios based on a deep learning model, which has the advantages of strong real-time performance, high robustness, and flexible hardware deployment.

[0032] In one embodiment, a vehicle radar and video data acquisition module is deployed at the edge node and combined with the MEC platform to achieve low-latency fusion of video data and radar data, thereby realizing real-time detection and behavior analysis of vehicles and other transportation tools, including: vehicle radar and video data acquisition; time synchronization; low-level data fusion; high-level semantic fusion; real-time processing and feedback.

[0033] In one embodiment, the vehicle perception target sensor data, including radar data and video data, is first collected using edge nodes installed in the traffic environment. The second step is to use MEC time synchronization to ensure that the timestamp of the video image data is consistent with the timestamp of the target radar data. The third step is to use the MEC low-level data fusion algorithm to perform preliminary fusion of deep learning-based target detection and radar data. The fourth step is to use the MEC high-level fusion algorithm to perform preliminary fusion of target features and behavior prediction after preliminary fusion. The fifth step is to use the NPU accelerator in the MEC for real-time target detection.

[0034] In one embodiment, vehicle radar and visual data collection utilizes edge nodes installed in a traffic environment to deploy vehicle radar sensors and visual sensors such as cameras to collect radar and visual data, including vehicle speed, acceleration, heading angle, length, width data, and video images to form raw data. The radar and visual data are sent to the vehicle perception target data processing module through the MEC platform; time synchronization utilizes the edge nodes of the MEC platform to deploy a clock synchronization server, namely the "time network protocol server (ntp)", to perform time synchronization, so that all video and radar data collection, processing, and fusion in the system are based on the same clock, and edge computing is used to perform frame alignment and timestamp marking on video images and radar data to form image frames with unified timestamps. Radar frame; Low-level data fusion step 1: Low-level data fusion uses the MEC platform to perform preliminary fusion of radar data and image data, and fuses radar data with visual data to preliminarily obtain the location, category, direction and other data of the image target; uses the deep learning target detection algorithm to detect the target vehicle in the image. The target detection algorithm is implemented on MEC, and the detection result is used as the basic data for the next step; Low-level target data matching uses the MEC platform to match the low-level target detection results with the radar target, and uses the visual target category and the visual target coordinates in the radar for preliminary matching. When the visual target category matches the radar target category and the visual target center coordinates are aligned with the radar target coordinate space, the low-level matching is judged to be successful.

[0035] In one embodiment, high-level semantic fusion is used for target trajectory and behavior prediction. This process is performed by the MEC platform, and specifically includes: based on the MEC platform, low-level matching results are combined with the physical information of the visual target, and target detection results are used as the basis for target feature extraction and feature fusion; high-level motion trajectory prediction uses physical information combined with target features to predict the target trajectory for target behavior judgment.

[0036] In one embodiment, a traffic target detection device is implemented using the following units, including: a radar perception unit, an image perception unit, an MEC module, an NPU hardware acceleration unit, a sensor data acquisition unit, a data preprocessing and synchronization unit, a low-level data fusion unit, a high-level fusion unit, and a real-time processing and feedback unit.

[0037] In one embodiment, the sensor data acquisition unit includes: a radar sensor and a camera.

[0038] In one embodiment, the data preprocessing and synchronization unit includes: a radar data filtering module and an image data filtering module.

[0039] In one embodiment, the low-level data fusion unit includes: a deep learning target detection model module, a radar data module, and an image data module.

[0040] In one embodiment, the high-level fusion unit includes: a target feature extraction module and a target behavior trajectory prediction module.

[0041] In one embodiment, the real-time processing and feedback unit includes: an NPU hardware unit, a target detection and behavior analysis module.

[0042] In one embodiment, the radar unit is deployed at the MEC edge node, and the image unit is also deployed at the MEC edge node.

[0043] In one embodiment, the working process of each functional module includes: the image perception unit uses a camera or other image sensor to collect and preprocess video images, and the data is processed by a real-time processing and feedback unit for real-time target detection and trajectory prediction.

[0044] In one embodiment, the radar sensing unit uses radar to collect vehicle information and preprocess data. The radar data is filtered and then transmitted to the MEC module.

[0045] In one embodiment, the MEC module performs data fusion on the received radar and image data, including: multi-frame fusion prediction of radar data, radar image target matching, and target detection and behavior analysis of the image through edge computing in MEC.

[0046] In one embodiment, the data preprocessing and synchronization unit processes information such as target speed, distance, angle, etc. obtained by the vehicle radar, as well as information such as target category and position, etc. obtained through images.

[0047] In one embodiment, the low-level data fusion performs preliminary fusion of target information acquired by the radar module and target information acquired by the image module, specifically by performing preliminary fusion through the speed, distance, and angle information of the radar target combined with the target category and position information in the image to obtain preliminary fusion data of the vehicle.

[0048] In one embodiment, the high-level fusion unit combines the preliminary fusion data of the vehicle to predict the trajectory of the target and obtain the target state information at the future moment. The specific process is as follows: based on the image data, the target feature extraction module of deep learning is used to extract the target feature data, and the category of the target is predicted through the prediction module; based on the radar data, the target trajectory is predicted in combination with the target feature prediction and the target category, and the target behavior is initially predicted; the target vehicle motion trajectory prediction, target vehicle category prediction, and target vehicle detection feature prediction are performed in the high-level fusion unit.

[0049] In one embodiment, the real-time processing and feedback unit performs real-time target vehicle detection, category prediction, and trajectory prediction through the NPU hardware unit based on the preliminary fusion results of the high-level fusion unit, providing decision information for road vehicle traffic management. The NPU hardware unit provides a low-latency, high-efficiency hardware foundation for target vehicle detection, accelerates deep learning network processing, and ensures real-time analysis of video data.

[0050] Compared with the prior art, the present invention has the following advantages: 1. The present invention combines the target detection algorithm based on deep learning with the MEC platform, uses the MEC platform to accelerate the target detection algorithm, and realizes the rapid output of the target detection results, thereby greatly improving the real-time and accuracy of target detection. 2. The present invention combines the target detection results with radar data for preliminary fusion, uses radar data as the verification of the authenticity of the target, combines images with convolutional neural networks for feature extraction, combines radar data such as position and angle with categories and features as parameter information, and performs preliminary fusion of vehicle trajectories, thereby realizing accurate prediction of vehicle behavior and providing a reference for traffic management. 3. The present invention combines traffic edge computing, deploys the target perception and detection algorithm on the MEC platform, does not require a large amount of computing resources, uses NPU hardware acceleration to achieve real-time target detection, and can quickly upgrade the version of the deployment algorithm through the edge computing platform, thereby improving the deployment efficiency of the algorithm. 4. The present invention combines the technical characteristics of 5G MEC, deploys the target detection algorithm on the MEC edge node, and quickly completes target detection through target data collection and fusion. It is particularly suitable for detection scenarios with strong real-time performance.

[0051] In other embodiments, the above method and device have the following additional technical features: edge perception and detection based on MEC, combined with deep learning and edge computing, radar and camera sensors deployed on MEC nodes are used to collect original images and radar data, MEC is used to pre-process, pre-align and pre-process the data, and the MEC platform is used for low-level data fusion, including preliminary target detection and preliminary target matching using the MEC target detection model; in high-level target fusion, a deep learning target feature extraction algorithm is used, combined with information such as target vehicles and target categories obtained by perception, to extract data features, and combined with the MEC multi-frame joint prediction algorithm, the target trajectory is fused through radar data and predicted trajectory to achieve target trajectory detection and behavior prediction.

[0052] In one embodiment, the MEC-based target vehicle detection method collects and preprocesses MEC data, uses the MEC data fusion module and the transfer learning method of the deep learning algorithm to perform target detection, uses radar target information combined with deep learning to predict target trajectories, and performs real-time target detection and feedback through the MEC NPU hardware unit.

[0053] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0054] According to an embodiment of the present invention, the present invention also provides an electronic device and a readable storage medium.

[0055] Electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the invention described and / or claimed herein.

[0056] The electronic device includes a computing unit, which can perform various appropriate actions and processes according to a computer program stored in a ROM or a computer program loaded from a storage unit into a RAM. In the RAM, various programs and data required for the operation of the electronic device can also be stored. The computing unit, ROM, and RAM are connected to each other via a bus. An I / O interface is also connected to the bus.

[0057] Multiple components in an electronic device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the electronic device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0058] The computing unit may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of computing units include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit performs the various methods and processes described above, such as traffic target detection methods. For example, in some embodiments, the traffic target detection method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on an electronic device via a ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the computing unit, one or more steps of the traffic target detection method described above may be performed. Alternatively, in other embodiments, the computing unit may be configured to perform the traffic target detection method in any other appropriate manner (e.g., by means of firmware).

[0059] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0060] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0061] In the context of the present invention, a readable storage medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A readable storage medium may be a machine-readable signal medium or a machine-readable storage medium. A readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. More specific examples of readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0062] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0063] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communications network). Examples of communications networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0064] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0065] It should be understood that the above-mentioned various forms of processes can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0066] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A traffic target detection method, characterized in that: include: Use vehicle radar sensors and video sensors installed on edge nodes to collect radar data and video data of target vehicles and traffic facilities; Accurately time-align video and radar data to ensure consistent timestamps of the data; Use image video data and radar data for preliminary matching, use visual data for target segmentation, match the segmented results with radar data, and complete the low-level fusion of the target vehicle; Using convolutional neural networks, the motion information in radar data and the image information in visual data are combined to extract and fuse target features, thus achieving semantic-level fusion. Real-time processing and feedback The above fused data is input into the NPU to perform real-time target detection and behavior analysis, obtain the target type, motion trajectory and behavior trajectory, and provide traffic guidance and prediction for the target vehicle.

2. The traffic target detection method according to claim 1, characterized in that: The radar data includes vehicle speed, acceleration, heading angle, length and width; the video data includes shape, color and texture; Radar data and video data are preprocessed through edge computing, including: denoising, filtering, signal enhancement of radar data, and denoising, enhancement, and feature extraction of image data.

3. The traffic target detection method according to claim 2, characterized in that: Use the edge nodes of the MEC platform to deploy NTP servers for time synchronization, so that the collection, processing and fusion of all video and radar data in the system are based on the same clock; - Use the MEC platform to perform spatial alignment of image frames or point clouds of radar data and video data to form a spatial coordinate mapping matrix, and complete the alignment of the same target in different spatial dimensions of the two data.

4. The traffic target detection method according to claim 3, characterized in that: Use YOLO, SSD or FasterRCNN deep learning target detection models to perform preliminary target detection on the spatial position of the target in the radar data and the image in the visual data to form the target category and regional position of the image; Based on the target position and category obtained in image target detection, the edge nodes of the MEC platform are used for target matching to find the corresponding targets in the radar data and form preliminary matching data.

5. The traffic target detection method according to claim 4, characterized in that: Based on the preliminary matching data, using the video target category and physical information in the radar data, data fusion is performed to extract target features and predict target behavior; The motion information in the radar data, the image information in the visual data, and the information on predicting target behavior are combined for high-level fusion to obtain the target's behavior trajectory and motion trajectory for traffic monitoring.

6. The traffic target detection method according to claim 5, characterized in that: Complete high-level semantic fusion at the MEC processing edge node, and send the fused data to the NPU for real-time target detection and behavior analysis. Use the NPU to perform real-time target detection and behavior analysis, and output the target type, motion trajectory, and behavior trajectory. Feedback the above data to the third-party receiving platform for traffic guidance and prediction.

7. The traffic target detection method according to claim 6, characterized in that: The MEC platform is used in conjunction with NPU hardware to accelerate the execution of deep learning target detection and classification algorithms.

8. An electronic device, characterized in that: include: at least one processor; A memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor so that the at least one processor can execute the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-6.

Images