Environment sensing system and method based on autonomous vehicle

Abstract

The invention relates to the technical field of automatic driving, in particular to an environment perception system and method based on an automatic driving vehicle, and the system comprises a visual processing module which carries out the first image preprocessing of a received image signal, and obtains image processing data generated through the first image preprocessing; the radar processing module is used for carrying out first data preprocessing on the collected environment data; the joint processing module is used for executing second data preprocessing on results output by the visual processing and the radar processing to generate environment sensing data; the automatic driving decision module controls vehicle response in combination with environment sensing data; the data compression module executes second image preprocessing on the image processing data to obtain image compression data; the detection area division module executes second data preprocessing on the image compression data to determine a detection area; and the radar target tracking module monitors and classifies objects in the detection area in real time and dynamically adjusts a predefined tracking strategy according to the position, speed and motion track parameters of the target object.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, in particular to an environment perception system and method based on an automatic driving vehicle. BACKGROUND

[0002] As an important part of intelligent transportation, automatic driving technology aims to realize the safe, efficient and intelligent operation of unmanned vehicles by utilizing advanced sensors and artificial intelligence technology. The environment perception system is one of the core technologies of automatic driving, which mainly collects information about the surrounding environment through various sensors such as cameras, radars and LiDARs, to help vehicles make driving decisions. In recent years, with the rapid development of deep learning and computer vision technology, automatic driving technology has attracted widespread attention, and relevant research and applications have been emerging. However, the reliability and accuracy of current automatic driving systems in complex environments still face challenges.

[0003] Traditional environment perception technology usually uses a single sensor for data collection, such as relying solely on cameras for visual recognition or using LiDAR for object detection. Although this method is effective under certain conditions, it is easily affected by weather, light changes and obstructions in complex and variable driving environments, resulting in inaccurate perception results. In addition, the traditional technology has limited capabilities in data processing and information fusion, often requiring a large amount of computing resources, making it difficult to meet the real-time needs of automatic driving. Therefore, improving the accuracy and real-time performance of environment perception has become a problem that needs to be solved in the current automatic driving field.

[0004] Although existing technologies have made some progress in the field of automatic driving, their limitations are obvious. First, the data from a single sensor cannot fully reflect the surrounding environment, resulting in insufficient perception accuracy in complex scenarios and increasing potential safety hazards. Second, the low data processing efficiency makes the system respond slowly in high dynamic environments, which cannot guarantee the safety and reliability of automatic driving. Third, traditional technology fails to effectively integrate information from multiple sensors, lacks intelligent decision support, and limits the overall performance of the automatic driving system. Therefore, there is an urgent need for an innovative environment perception and testing system that improves the performance of existing technologies through multi-sensor fusion and intelligent algorithms to address the shortcomings of traditional technology. SUMMARY

[0005] To achieve the above-mentioned needs, the present application provides an environment perception system and method based on an automatic driving vehicle, which improves the environment perception performance of the vehicle in different scenarios through multi-sensor fusion and intelligent algorithms, ensuring the safety and reliability of automatic driving. It remedies the potential safety hazards caused by insufficient perception accuracy in complex scenarios and the problem of slow response in high dynamic environments caused by low data processing efficiency in traditional technology.

[0006] To achieve the above-mentioned object of the application, the application adopts the following technical solutions:

[0007] In a first aspect, the application provides an environment perception system based on an autonomous vehicle, comprising a vision processing module, a radar processing module, a joint processing module, an autonomous driving decision module, a data compression module, a detection region division module, and a radar target tracking module.

[0008] The vision processing module is configured to receive image signals from a vehicle-mounted camera, perform first image preprocessing on the image signals, and obtain image processing data generated by the first image preprocessing.

[0009] The radar processing module is configured to perform first data preprocessing on environment data collected by a radar sensor.

[0010] The joint processing module is configured to perform second data preprocessing on the results output by the vision processing module and the radar processing module based on a multi-sensor fusion algorithm, and generate environment perception data.

[0011] The autonomous driving decision module is configured to generate control instructions in combination with the environment perception data output by the joint processing module, and control the response of the vehicle to target objects in different driving scenarios.

[0012] The data compression module is configured to perform second image preprocessing on the image processing data, and obtain image compression data generated by the second image preprocessing.

[0013] The detection region division module is configured to perform third image preprocessing on the image compression data, and perform second data preprocessing on the image compression data after the third image preprocessing, and determine a detection region.

[0014] The radar target tracking module is configured to perform real-time monitoring and classification on objects in the detection region based on a Kalman filtering algorithm, and dynamically adjust a pre-defined tracking strategy according to the position, speed, and motion trajectory parameters of the target objects.

[0015] The first image preprocessing includes adaptive adjustment of image resolution.

[0016] The second image preprocessing includes compression of image data, removal of noise and redundant information in the image.

[0017] The third image preprocessing includes division of image regions.

[0018] The first data processing includes detection of distance, speed, and position information of objects around the vehicle.

[0019] The second data processing comprises: using a multi-source data fusion algorithm to perform fusion processing on the data, and converting the data after the fusion processing into environment perception data and forwarding the environment perception data to an automatic driving decision module.

[0020] The predefined tracking strategy comprises: in the process of vehicle driving, taking the relative motion state change of the vehicle speed and the target object as the basis, and dynamically adjusting the tracking accuracy and the update frequency according to the speed and acceleration parameters of the vehicle when the vehicle is in the acceleration or deceleration state.

[0021] Optionally, the visual processing module comprises: a vehicle-mounted camera installed in the front and surrounding area of the vehicle, and a collection sub-module, an image processing sub-module and an image data output sub-module connected with the vehicle-mounted camera.

[0022] The vehicle-mounted camera continuously collects image signals of the environment around the vehicle at a specific frame rate.

[0023] The collection sub-module is configured to collect images through the vehicle-mounted camera to obtain image signals.

[0024] The image processing sub-module is configured to use an image processing algorithm to perform parallel processing on high-resolution images and low-resolution images in the image signals, and obtain high-resolution image processing data and low-resolution image processing data, respectively.

[0025] The image processing algorithm comprises a visual recognition method based on a convolutional neural network (CNN) and a fast algorithm based on edge detection.

[0026] The high-resolution image processing data comprises accurate position information of a long-distance object and object category information.

[0027] The low-resolution image processing data comprises edge features of a close-range object.

[0028] The image data output sub-module is configured to output the image processing data obtained by the image processing sub-module in the form of feature maps and position information, and the output data is composed of feature vectors and object detection boxes, and the format is .

[0029] Wherein, x and y are the coordinates of the upper left corner of the object, and w and h are the width and height of the object.

[0030] Optionally, the image processing sub-module comprises a first image processing unit configured to use a convolutional layer of a deep convolutional network to extract high-order features of the image and identify long-distance obstacles and road condition information for the high-resolution image, and obtain a set of deep features of the high-resolution image after processing by a plurality of convolutional and pooling layers.

[0031] The second image processing unit is configured to process the close-range environment around the vehicle, perceive the adjacent vehicles, pedestrians and other moving objects, based on a fast algorithm of edge detection for the low-resolution image.

[0032] Optionally, the radar processing module comprises:

[0033] The radar sensor is configured to transmit electromagnetic waves and receive echo signals, and detect the distance, speed and position information of the objects in front of and around the vehicle.

[0034] The radar signal input sub-module is configured to receive the echo signals of the radar sensor, process the echo signals according to the radar algorithm, and generate preliminary radar scanning data; the radar scanning data comprises the distance, angle and intensity information of the objects.

[0035] The visual data fusion sub-module is configured to receive the image processing data output by the image data output sub-module, and detect and identify the objects by fusing the visual data.

[0036] The radar algorithm application sub-module is configured to analyze and detect the obstacles in the detection area based on the radar algorithm, and determine the position, speed and size of the obstacles.

[0037] Optionally, the joint processing module comprises:

[0038] The multi-source data fusion sub-module is configured to receive the output data of the radar processing module and the visual processing module, and perform fusion processing.

[0039] The environment perception data generation sub-module is configured to convert the fusion-processed data into environment perception data, and output the environment perception data to the automatic driving decision module; the environment perception data comprises the distance, angle, speed and position information of the objects.

[0040] Optionally, the data compression module comprises:

[0041] The compression processing sub-module is configured to receive the high-resolution image data from the visual processing module, and perform compression based on a specific compression algorithm.

[0042] The compressed data output sub-module is configured to output the compressed data to the radar processing module; the compressed data comprises the image compression data of the front and surrounding of the vehicle.

[0043] Optionally, the detection area division module comprises a background segmentation sub-module and a region division sub-module.

[0044] The background segmentation submodule is configured to binarize the image compression data, divide pixels in the image into foreground pixels and background pixels, and after binarization, eliminate noise in the binarized image through morphological opening operation, separate the background part from the foreground object in the binarized image, and obtain a relatively smooth and noise-free foreground image.

[0045] The region division submodule is configured to distinguish different foreground regions by image segmentation processing according to the connection relationship between the foreground pixels through a region labeling method based on a connected domain.

[0046] The binarization of the image compression data and the division of the pixels in the image into foreground pixels and background pixels include: setting an input grayscale image as , and performing binarization operation according to a set threshold T, as follows:

[0047]

[0048] In the formula, d represents the distance between pixels, and D is a connectivity threshold. is a binarized image, and a pixel value of 1 represents a foreground pixel, and a pixel value of 0 represents a background pixel.

[0049] The elimination of noise in the binarized image through morphological opening operation includes: setting a structure element as S, and the formula of the opening operation is as follows:

[0050]

[0051] In the formula, d represents the distance between pixels, and D is a connectivity threshold. represents an erosion operation, represents an expansion operation, is a binarized image after the opening operation.

[0052] Optionally, the region division submodule includes:

[0053] The searching unit is configured to search all connected foreground pixel clusters in the binarized image, and label each pixel cluster. Assuming that a pixel cluster is , all pixels inside the pixel cluster satisfy the connectivity condition, that is:

[0054]

[0055] In the formula, d represents the distance between pixels, and D is a connectivity threshold.

[0056] The smoothing detection unit is configured to effectively divide different regions in the foreground pixels into a plurality of connected detection regions, and use morphological closing operation to smooth the boundaries of the detection regions.

[0057] The formula of the closing operation is as follows:

[0058] .

[0059] Optionally, the radar target tracking module comprises:

[0060] a target identification and tracking sub-module, configured to receive target object information from the detection region division module in a target identification and tracking stage, and combine the radar echo signal to calculate the speed and position of the target object in real time;

[0061] wherein the relative speed of the target object is calculated by the following formula:

[0062]

[0063] wherein v is the relative speed of the target object, is the frequency of the radar signal, and c is the speed of light; is the measured Doppler frequency; is the frequency difference between the received echo signal and the transmitted signal;

[0064] the spatial position of the target object is located by the following formula:

[0065]

[0066] wherein B is the signal bandwidth of the radar transmission, and d is the distance between the target and the radar;

[0067] a radar target tracking sub-module, configured to track the trajectory of the target object by a Kalman filter to determine the motion state estimation value of the object, as follows:

[0068]

[0069]

[0070] wherein is the state vector of the target, including the position x, y, the speed , information, F is the state transition matrix, is the process noise, represents the measurement value at each time, H is the observation matrix, is the measurement noise;

[0071] an obstacle classification sub-module, configured to judge a moving object or a stationary object according to the motion state of the target object in an obstacle classification stage; and for a moving target, classify it according to the speed and motion characteristics.

[0072] In a second aspect, the present application provides an environment perception method based on an automatic driving vehicle, which comprises: receiving an image signal from a vehicle-mounted camera, performing first image preprocessing on the image signal to obtain image processing data generated by the first image preprocessing.

[0073] performing first data preprocessing on the environment data collected by the radar sensor;

[0074] performing second data preprocessing on the image processing data and the result of the first data preprocessing based on a multi-sensor fusion algorithm to generate environment perception data;

[0075] generating control instructions in combination with the environment perception data to control the response of the vehicle to the target object in different driving scenarios;

[0076] performing second image preprocessing on the image processing data to obtain image compression data generated by the second image preprocessing image processing data;

[0077] performing third image preprocessing on the image compression data and performing second data preprocessing on the image compression data after the third image preprocessing to determine a detection area;

[0078] based on a Kalman filter algorithm, real-time monitoring and classification of objects in the detection area, and dynamic adjustment of the pre-defined tracking strategy according to the position, speed and motion trajectory parameters of the target object;

[0079] The first image preprocessing includes adaptive adjustment of image resolution.

[0080] The second image preprocessing includes compression of image data to remove noise and redundant information in the image.

[0081] The third image preprocessing includes dividing the image area.

[0082] The first data processing includes detecting the distance, speed and position information of the objects around the vehicle.

[0083] The second data processing includes using a multi-source data fusion algorithm to perform fusion processing on the data, converting the fusion-processed data into environment perception data and forwarding it to the automatic driving decision module.

[0084] The pre-defined tracking strategy includes that during the driving of the vehicle, the relative motion state change of the vehicle speed and the target object is proportional, and according to the speed and acceleration parameters of the vehicle, when the vehicle is in an acceleration or deceleration state, the tracking accuracy and update frequency are dynamically adjusted.

[0085] Compared with the closest prior art, the present application has the following beneficial effects:

[0086] The application provides an environment perception system and method based on an automatic driving vehicle, which can realize real-time and comprehensive perception of the surrounding environment through a multi-sensor fusion technology, effectively identify obstacles and pedestrians, and greatly reduce the risk of traffic accidents, thereby improving the safety of the automatic driving vehicle.

[0087] The integrated visual processing module and radar processing module of the system can efficiently collect and process data in various complex environments, ensure accurate detection of objects at different distances, and improve the accuracy and reliability of environment perception, and can flexibly cope with different driving scenarios, including urban roads, highways and adverse weather conditions, and through dynamic adjustment of the processing mode, ensure efficient operation in changing environments.

[0088] The perception data obtained by the application provides a reliable basis for the automatic driving decision module, promotes the vehicle to make quick and accurate decisions in various traffic conditions, and improves the intelligent level of automatic driving.

[0089] The data compression module is provided in the application, and the design concept of the data compression module effectively reduces the burden of data storage and transmission, improves the operation efficiency of the system, reduces the demand for computing resources, and enables the system to maintain efficient work in the case of limited resources.

[0090] In addition, the application provides an environment perception system and method based on an automatic driving vehicle, which can accurately track the trajectory of moving objects through dynamic monitoring of the radar target tracking module, ensure timely identification and response to potential dangers, and improve the reaction capability of the automatic driving system.

[0091] In the manual driving state, the system provides auxiliary information of the surrounding environment, improves the driver's perception ability of the environment, makes the driving more safe and convenient, and optimizes the user experience. BRIEF DESCRIPTION OF DRAWINGS

[0092] In order to more clearly illustrate the specific embodiments of the application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual proportion.

[0093] Figure 1 is a system architecture schematic diagram of an environment perception system based on an automatic driving vehicle provided by the application;

[0094] Figure 2 is a structural block diagram of an environment perception system based on an automatic driving vehicle provided by the application;

[0095] Figure 3It is a kind of environment perception method flow chart based on automatic driving vehicle provided by the present application. DETAILED DESCRIPTION

[0096] The embodiments of the technical solutions of the present application will be described in detail below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present application, therefore, only as an example, and cannot limit the protection scope of the present application.

[0097] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should be the usual meaning understood by the skilled person in the field of the present application.

[0098] The present application provides an environment perception system and method based on automatic driving vehicle. It aims to achieve accurate perception and real-time data processing in complex driving environment through the collaborative work of multiple modules such as visual processing, radar processing and data fusion. The system integrates advanced image processing and radar algorithms to achieve comprehensive perception of the environment around the vehicle, and processes and fuses the data of different sensors in real time to ensure that the automatic driving vehicle can operate safely and reliably in various complex scenarios. The embodiments of the present application will be described below with reference to the drawings.

[0099] In one embodiment, the environment perception system based on automatic driving vehicle provided by the embodiment of the present application, Figure 1 As shown, the overall architecture mainly consists of visual processing module, radar processing module, joint processing module, data compression module, detection region division module, radar target tracking module and other auxiliary modules. Through the collaborative work of these modules, the system can realize the whole process from the collection and processing of environmental data to the final generation of perception information for automatic driving control. The system is not only suitable for urban road scenes, but also can be used for highway, complex terrain and other driving scenes.

[0100] As Figure 2 As shown, the system specifically includes: visual processing module 110, radar processing module 120, joint processing module 130, automatic driving decision module 140, data compression module 150, detection region division module 160 and radar target tracking module 170;

[0101] The specific functions are as follows: the visual processing module 110 is used to receive image signals from the vehicle-mounted camera, perform first image preprocessing on the image signals, and obtain image processing data generated by the first image preprocessing of the image signals;

[0102] The radar processing module 120 is used to perform first data preprocessing on the environmental data collected by the radar sensor;

[0103] The joint processing module 130 is configured to perform second data preprocessing on the results output by the visual processing module and the radar processing module based on a multi-sensor fusion algorithm, and generate environment perception data.

[0104] The automatic driving decision module 140 is configured to generate a control instruction in combination with the environment perception data output by the joint processing module, and control the response of the vehicle to the target object in different driving scenarios.

[0105] The data compression module 150 is configured to perform second image preprocessing on the image processing data to obtain image compression data generated by the second image preprocessing image processing data.

[0106] The detection region division module 160 is configured to perform third image preprocessing on the image compression data, and perform second data preprocessing on the image compression data after the third image preprocessing to determine a detection region.

[0107] The radar target tracking module 170 is configured to perform real-time monitoring and classification on the objects in the detection region based on a Kalman filter algorithm, and dynamically adjust a pre-defined tracking strategy according to the position, speed and motion trajectory parameters of the target object.

[0108] The first image preprocessing includes adaptive adjustment of image resolution.

[0109] The second image preprocessing includes compressing the image data to remove noise and redundant information in the image.

[0110] The third image preprocessing includes dividing the image region.

[0111] The first data processing includes detecting the distance, speed and position information of the objects around the vehicle.

[0112] The second data processing includes using a multi-source data fusion algorithm to perform fusion processing on the data, and converting the fusion-processed data into environment perception data and forwarding the environment perception data to the automatic driving decision module.

[0113] The pre-defined tracking strategy includes that during the driving of the vehicle, the relative motion state change of the vehicle speed and the target object is proportional, and according to the speed and acceleration parameters of the vehicle, when the vehicle is in an acceleration or deceleration state, the tracking accuracy and update frequency are dynamically adjusted.

[0114] It can be understood that the dynamic adjustment of the tracking strategy in the present application is one of the key functions of the radar target tracking module. During the driving of the vehicle, the relative motion state of the target will change with the change of the speed. Therefore, the module will dynamically adjust the tracking accuracy and update frequency according to the speed, acceleration and other parameters of the vehicle. When the vehicle is in the state of acceleration or deceleration, the module will increase the update frequency of target tracking to ensure timely response. For example, when the vehicle accelerates from a stationary state to , the tracking update frequency of the system can be dynamically adjusted according to the acceleration of the vehicle. Assuming that the update period of the radar data is , the target tracking module will adjust the update period to:

[0115]

[0116] In the formula, k is the adjustment coefficient, indicating the influence of acceleration on the update period. When the acceleration is large, the update period is shortened, and the system can track the target object more timely.

[0117] In addition, the above-mentioned system provided in the specific embodiment of the present application also includes other auxiliary modules; for example, the vehicle state perception module, which can determine the real-time state of the vehicle through the motion state of the vehicle in the system, and the system will dynamically adjust the input mode of image acquisition according to the state of the vehicle. Specifically, it is divided into the following two states:

[0118] First, the automatic driving state: when the vehicle is in the automatic driving state, the system will preferentially process the high-resolution image processing data in front of the vehicle to ensure accurate perception of the environment in front. In this state, the system mainly relies on real-time data of vision and radar sensors to make automatic driving decisions.

[0119] Second, the manual driving state: in the manual driving state, the system will receive all image data in front of the vehicle and around the vehicle, including high-resolution image data and compressed image data. These data will be provided as auxiliary information to the driver to help him understand the environment around the vehicle in real time.

[0120] For another example, the data fusion and output module, through the data fusion algorithm of multiple sensors, finally generates the environment perception data. These data not only include the distance, angle, speed, size and other information of the object, but also output the motion trajectory of the object through the radar target tracking module. After data fusion processing, the environment perception result generated by the system can be fed back to the automatic driving decision module of the vehicle in real time to ensure that the vehicle can make accurate responses in various driving scenarios. The system is particularly suitable for complex roads, intersections, tunnels and adverse weather conditions, and through the optimized sensor fusion technology, the environment perception ability of the automatic driving vehicle can be significantly improved, thereby improving the safety and stability of driving.

[0121] In the above embodiment, the visual processing module 110 serves as the front-end module of the system, responsible for receiving image signals from the vehicle-mounted camera and performing resolution adaptive processing on the visual image through advanced image processing algorithms, thereby ensuring the provision of high-precision perception data in a complex and changing driving environment. The cameras are arranged at different positions of the vehicle, including the front, side, and rear of the vehicle, to achieve all-round environmental perception. The image resolutions processed by this module mainly include high-resolution and low-resolution images, which are used for long-range and close-range environmental perception, respectively. To ensure the stability and robustness of the algorithm, the visual processing module can continue to work under various lighting conditions.

[0122] The visual processing module 110 includes: a vehicle-mounted camera installed in the front and surrounding areas of the vehicle, and a collection submodule, an image processing submodule and an image data output submodule connected to the vehicle-mounted camera;

[0123] The vehicle-mounted camera continuously collects image signals of the vehicle's surrounding environment at a specific frame rate;

[0124] The acquisition submodule is used to acquire images through the vehicle-mounted camera and obtain image signals;

[0125] The image processing submodule is used to process the high-resolution and low-resolution images in the image signal in parallel using an image processing algorithm to obtain high-resolution image processing data and low-resolution image processing data respectively;

[0126] The image processing algorithm includes: a visual recognition method based on convolutional neural network (CNN) and a fast algorithm based on edge detection;

[0127] The high-resolution image processing data includes: precise location information of distant objects and object category information;

[0128] The low-resolution image processing data includes: edge features of close-range objects;

[0129] The image data output submodule is used to output the image processing data obtained by the image processing submodule in the form of feature maps and position information. The output data consists of feature vectors and object detection frames in the format of ;

[0130] Among them, x, y are the coordinates of the upper left corner of the object; w, h are the width and height of the object.

[0131] The image processing submodule includes: a first image processing unit for identifying long-distance obstacles and road condition information by extracting high-order features of the high-resolution image using the convolution layer of a deep convolutional network; after processing through several convolution and pooling layers, a set of deep features of the high-resolution image is obtained;

[0132] A second image processing unit is configured to process the close-range environment around the vehicle based on a fast algorithm of edge detection for the low-resolution image, and to perceive proximate vehicles, pedestrians and other moving objects.

[0133] In particular, the workflow of the visual processing module 110 is as follows:

[0134] First, the image acquisition and input stage acquires data through the cameras installed on the front and around the vehicle. The cameras continuously capture the environmental images around the vehicle at a specific frame rate, and the acquired image signals are transmitted into the system through a standard data interface. In the image processing stage, the module processes the high-resolution and low-resolution images in parallel by applying different image processing algorithms.

[0135] The specific image processing algorithm adopts a visual recognition method based on a convolutional neural network (CNN). For high-resolution images, the system mainly uses a deep convolutional network. The convolutional layer extracts high-order features of the image to recognize distant obstacles and road conditions. Let the input image be (high resolution), the convolutional network first performs a first layer of convolution operation:

[0136]

[0137] wherein, is the convolution kernel, is the bias term, is the activation function (such as ReLU), ∗ represents the convolution operation, is the output feature map. After each convolution layer, the system further reduces the size of the feature map through a pooling layer (such as max pooling), thereby reducing the computational load and enhancing the robustness of the features.

[0138] Subsequently, after processing through several convolution and pooling layers, the system finally obtains a set of deep features of high-resolution images. These features are used to recognize distant obstacles and complex road details, such as other vehicles on the road, distant roadblocks, traffic signs, etc. Assuming that there is an output layer representing the final classification result, combined with the Softmax activation function, object classification can be performed through the following formula:

[0139]

[0140] wherein, represents the probability that the object belongs to class CCC, is the score of each class j.

[0141] For low-resolution images, the processing is mainly for the close-range environment around the vehicle. Low-resolution images are usually used to perceive adjacent vehicles, pedestrians and other moving objects. In processing, the visual processing module adopts a fast algorithm based on edge detection. First, the Sobel operator is applied for gradient calculation. Let the low-resolution image be Then the gradients and are respectively:

[0142] ,

[0143] The edges of the image are calculated by the gradient direction and amplitude:

[0144]

[0145] In this way, the edge information of the objects around the vehicle can be quickly detected, helping to identify adjacent obstacles.

[0146] In terms of image data output, the processed image data is output in the form of feature maps and position information. High-resolution image processing data contains accurate position information of distant objects, object categories and other details, while low-resolution image processing data mainly contains edge features of close-range objects. The final output data consists of feature vectors and object detection boxes, and the format is where x, y are the coordinates of the upper left corner of the object, and w, h are the width and height of the object.

[0147] In the above embodiment, the radar processing module 120 comprises:

[0148] a radar sensor for sending electromagnetic waves and receiving echo signals, detecting the distance, speed and position information of objects in front of and around the vehicle;

[0149] a radar signal input sub-module for receiving echo signals of the radar sensor and processing the echo signals according to the radar algorithm to generate preliminary radar scan data; the radar scan data includes distance, angle and intensity information of the object;

[0150] a visual data fusion sub-module for receiving image processing data output by the image data output sub-module, detecting and identifying objects through fusion with visual data;

[0151] a radar algorithm application sub-module for analyzing and detecting obstacles in the detection area based on the radar algorithm to determine the position, speed and size of the obstacles.

[0152] In particular, the radar processing module 120 processes the environmental data collected by the radar sensor. The radar sensor detects the distance, speed, and position of objects in front of and around the vehicle by emitting electromagnetic waves and receiving echoes. Specifically, the workflow of the radar processing module is as follows:

[0153] Radar data reception: The radar processing module receives the echo signals from the radar sensor and processes them according to the radar algorithm to generate preliminary radar scan data. These data include the distance, angle, and intensity of the detected objects.

[0154] Fusion with visual data: The radar processing module also receives image processing data output by the visual processing module, especially high-resolution image processing data and low-resolution image compression data. Through the fusion of visual data, the radar processing module can more accurately detect and identify objects.

[0155] Radar algorithm application: The radar processing module analyzes the obstacles in the detection area based on the radar algorithm to determine their position, speed, and size. This module is particularly suitable for target recognition in complex environments, such as insufficient light and harsh weather conditions.

[0156] In the above embodiment, the joint processing module 130 includes:

[0157] Multi-source data fusion sub-module: receives output data from the radar processing module and the visual processing module and performs fusion processing;

[0158] Environment perception data generation sub-module: converts the fusion-processed data into environment perception data and outputs it to the autonomous driving decision module; the environment perception data includes the distance, angle, speed, and position information of the objects.

[0159] The joint processing module 130 is the core processing unit of the system, responsible for fusing data from different sensors to generate final environment perception data. The working principle of this module is based on a multi-sensor fusion algorithm, which combines the results of visual processing and radar processing to improve the accuracy and reliability of environment perception. Specifically, the workflow of the joint processing module is as follows:

[0160] Multi-source data fusion: The joint processing module 130 receives output data from the radar processing module and the visual processing module and performs fusion processing on these data. The multi-source data fusion algorithm can integrate the advantages of different sensors to improve the recognition rate and detection accuracy of obstacles.

[0161] Specifically, it includes the following two parts:

[0162] First, environmental perception data generation: the joint processing module converts the fused data into final environmental perception results and outputs them to the vehicle's decision-making module. These environmental perception data include object distance, angle, speed, position, and other information, providing reliable basis for the control decision of autonomous vehicles.

[0163] Second, multi-scenario handling: this module performs well in different scenarios, especially in complex urban traffic, tunnels, bridges, and harsh weather conditions. Multi-sensor fusion technology can greatly improve the vehicle's perception ability.

[0164] In the above embodiment, the data compression module 150 includes:

[0165] The compression processing submodule is used to receive high-resolution image data from the visual processing module and compress it based on a specific compression algorithm.

[0166] The compressed data output submodule is used to output compressed data to the radar processing module; wherein the compressed data includes image compression data of the front and surrounding of the vehicle.

[0167] The data compression module 150 is used to reduce the storage and transmission load of high-resolution image data output by the visual processing module. By compressing image data, this module can significantly improve the efficiency of the system. Its workflow is as follows:

[0168] Compression processing: the data compression module first receives high-resolution image data from the visual processing module and compresses it based on a specific compression algorithm. The compressed data volume is greatly reduced, but still maintains high image quality, meeting the needs of subsequent processing.

[0169] Compression data output: after compression processing, the data compression module outputs compressed data to the radar processing module for further radar data processing. The compressed data includes image compression data of the front and surrounding of the vehicle.

[0170] In the above embodiment, the core function of the detection region division module 160 is to divide the pre-processed image data into regions so that the system can identify and detect target objects in different regions. This module uses a background segmentation technique based on morphological image processing to remove noise and redundant information in the image, making the foreground target more clearly presented. The entire process consists of two main steps: background segmentation and region division.

[0171] First, the core goal of background segmentation is to separate the background part from the foreground objects in the image. In morphological processing, commonly used operations include dilation, erosion, opening operation and closing operation.

[0172] Specifically, the detection region division module 160 is composed of a background segmentation sub-module and a region division sub-module;

[0173] The background segmentation sub-module is configured to binarize the image compression data, divide the pixels in the image into foreground pixels and background pixels, and after the binarization, eliminate the noise in the binarized image through morphological opening operation, separate the background part from the foreground object in the binarized image, and obtain a relatively smooth and noise-free foreground image.

[0174] The region division sub-module is configured to divide different foreground regions by image segmentation processing according to the connection relationship between the foreground pixels through a region labeling method based on connected domains.

[0175] The binarization of the image compression data and the division of the pixels in the image into foreground pixels and background pixels include: assuming that the input gray image is , and the binarization operation is performed according to a set threshold T, as follows:

[0176]

[0177] In the formula, the binarized image, the pixel value of 1 represents a foreground pixel, and the pixel value of 0 represents a background pixel.

[0178] The elimination of the noise in the binarized image through the morphological opening operation includes: assuming that the structure element is S, and the formula of the opening operation is:

[0179]

[0180] In the formula, represents an erosion operation, represents an expansion operation, is the binarized image after the opening operation. The erosion can remove small noise on the edge of the object, and the expansion can repair the small cracks caused by the erosion, so that the foreground target is more prominent. After the opening operation is completed, the system will obtain a relatively smooth and noise-free foreground image.

[0181] Next, the system enters the region division stage. In this stage, the system further divides the image through a region labeling method based on connected domains. The connected domain algorithm can distinguish different foreground regions according to the connection relationship between the foreground pixels.

[0182] Specifically, the region division sub-module includes:

[0183] The searching unit is configured to search all connected foreground pixel clusters in the binarized image and label each pixel cluster. Assuming that a pixel cluster is , all the pixels inside the pixel cluster satisfy the connectivity condition, that is:

[0184]

[0185] where d represents the distance between pixels, and D is the connectivity threshold value;

[0186] The smoothing detection unit is used to effectively divide different regions in the foreground pixels into several connected detection regions; the boundary of the detection region is smoothed using morphological closing operation;

[0187] After the region division is completed, the system uses morphological closing operation to further smooth the boundary of the detection region. The closing operation is composed of an inflation operation followed by an erosion operation, which can fill small holes in the foreground object and improve the boundary continuity of the object.

[0188] The formula of the closing operation is:

[0189] .

[0190] After the above processing, the divided detection regions have smoother edges, ensuring that the subsequent radar processing module and joint processing module can more accurately identify and track target objects in these regions.

[0191] In one embodiment, it is assumed that there is a moving car and two pedestrians in front of the vehicle. Through background segmentation, the detection region division module first removes the background region and only keeps the vehicle and pedestrians as foreground. Then, the system divides the vehicle and pedestrians in the foreground into different detection regions through the connected component algorithm. For each detection region, the system marks the boundary of the region and passes this information to the subsequent processing module to ensure that the vehicle's perception system can identify these different targets and perform further tracking and analysis.

[0192] Finally, the region information generated by the detection region division module is output in the form of a rectangular box, and the coordinates of each rectangular box are defined as (x, y, w, h), where x and y are the coordinates of the upper left corner of the rectangular box, and w and h are the width and height of the rectangular box, respectively. These detection region information provides high-precision region division results for subsequent radar processing and joint processing, enabling the system to more accurately identify and track target objects around the vehicle in complex driving scenarios.

[0193] In the above embodiment, the radar target tracking module 170 continuously tracks and classifies target objects around the vehicle through accurate radar data processing and Kalman filtering algorithm. It not only can calculate the speed and position of the target in real time, but also can dynamically adjust the tracking strategy to ensure that the system can timely identify and respond to potential dangers in complex driving environments, thereby ensuring driving safety.

[0194] The radar target tracking module 170 comprises:

[0195] A target identification and tracking sub-module, configured to, in a target identification and tracking stage, receive target object information from the detection region division module, and combine radar echo signals to calculate the speed and position of the target object in real time;

[0196] First, in the target identification and tracking stage, the radar target tracking module receives target object information from the detection region division module, and combines radar echo signals to accurately position each target. Radar signals usually use pulse Doppler radar or FMCW (continuous wave frequency modulation) radar technology, which can simultaneously detect the distance and relative speed of the target object. Assuming that the frequency of the radar signal is , the relative speed of the target object is v, and according to the Doppler effect, the echo frequency received by the radar changes.

[0197] wherein the relative speed of the target object is calculated by the following formula:

[0198]

[0199] In the formula, v is the relative speed of the target object, is the frequency of the radar signal, and c is the speed of light; is the measured Doppler frequency; is the frequency difference between the received echo signal and the transmitted signal;

[0200] The spatial position of the target object is positioned by the following formula:

[0201]

[0202] In the formula, B is the signal bandwidth of the radar transmission; d is the distance between the target and the radar; through the formula, the module can accurately calculate the distance between the target and the radar, thereby accurately positioning the spatial position of the target object.

[0203] After obtaining the speed and position of the target, the radar target tracking module further tracks the trajectory of the target object through a Kalman filter. The Kalman filter is a recursive algorithm for processing noisy data, which can update the state estimate of the target in multiple measurements. Assuming that the state vector of the target is, containing position x, y, speed, and other information.

[0204] A radar target tracking sub-module, configured to track the trajectory of the target object through a Kalman filter to determine the motion state estimate of the object, as follows:

[0205]

[0206]

[0207] wherein, is the state vector of the target, including position x, y, velocity , information, F is the state transition matrix, is the process noise, represents the measurement value at each time, H is the observation matrix, is the measurement noise; through the prediction and update steps of the Kalman filter, the system can continuously correct the position and velocity estimates of the target object, so that the tracking process has high precision and robustness.

[0208] The obstacle classification submodule is used to judge a moving object or a stationary object according to the motion state of the target object in the obstacle classification stage; for a moving target, the moving target is classified according to the speed and motion characteristics.

[0209] In the obstacle classification stage, the radar target tracking module judges the category of the target object according to the motion state of the target object. The system judges whether the target object is a moving object or a stationary object by analyzing the trajectory form of the target object. For a moving target, the module further classifies the moving target according to the speed and motion characteristics of the moving target. For example, pedestrians usually have a low-speed and randomly changing motion trajectory, while vehicles have a faster motion speed and a stable driving trajectory. Assuming that the system needs to distinguish pedestrians from vehicles, a classification threshold value can be set according to the speed range of the target object; if the speed v of a target satisfies:

[0210]

[0211] the target is classified as a pedestrian; otherwise, if the speed of the target is greater than the threshold value, the target is classified as a vehicle. In addition, the system can also combine the size information of the target object to further enhance the classification accuracy. Through such multi-dimensional analysis, the radar target tracking module can accurately classify the surrounding obstacles.

[0212] Based on the same inventive concept, the embodiment of the present application also provides an environment perception method based on an autonomous vehicle. The implementation scheme of the problem solving provided by the method is similar to the implementation scheme described in the above-mentioned embodiment system, so the specific limitations in one or more environment perception methods based on an autonomous vehicle provided below can refer to the limitations of the environment perception system based on an autonomous vehicle described above, and will not be repeated here.

[0213] In one embodiment, please refer to Figure 3 , Figure 3An environment perception method based on an autonomous vehicle is provided for Embodiment 1 of the present application, and the specific steps of the method include:

[0214] S101 receives an image signal from a vehicle-mounted camera, performs first image preprocessing on the image signal, and obtains image processing data generated by the first image preprocessing;

[0215] S102 performs first data preprocessing on the environment data collected by the radar sensor;

[0216] S103, based on a multi-sensor fusion algorithm, performs second data preprocessing on the image processing data and the result of the first data preprocessing, and generates environment perception data;

[0217] S104 generates a control instruction in combination with the environment perception data, and controls the response of the vehicle to the target object in different driving scenarios;

[0218] S105 performs second image preprocessing on the image processing data, and obtains image compression data generated by the second image preprocessing;

[0219] S106 performs third image preprocessing on the image compression data, and performs second data preprocessing on the image compression data after the third image preprocessing to determine a detection area;

[0220] S107, based on a Kalman filter algorithm, performs real-time monitoring and classification on the objects in the detection area, and dynamically adjusts a pre-defined tracking strategy according to the position, speed and motion trajectory parameters of the target object;

[0221] The first image preprocessing includes adaptive adjustment of image resolution;

[0222] The second image preprocessing includes compressing the image data and removing noise and redundant information in the image;

[0223] The third image preprocessing includes dividing the image area;

[0224] The first data processing includes detecting the distance, speed and position information of the objects around the vehicle;

[0225] The second data processing includes using a multi-source data fusion algorithm to perform fusion processing on the data, converting the fusion-processed data into environment perception data, and forwarding the environment perception data to an autonomous driving decision module;

[0226] The pre-defined tracking strategy includes that, during the driving of the vehicle, the relative motion state change of the vehicle speed and the target object is proportional, and according to the speed and acceleration parameters of the vehicle, when the vehicle is in an acceleration or deceleration state, the tracking accuracy and update frequency are dynamically adjusted.

[0227] It should be understood that, although the steps in the flowcharts involved in the embodiments described above are shown in a sequence as indicated by arrows, the steps are not necessarily executed in the order as indicated by the arrows. Unless otherwise specified herein, the execution of the steps is not necessarily limited in sequence, and the steps can be executed in other sequences. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution of the steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least some of the other steps or the steps or stages in the other steps.

[0228] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0229] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce an apparatus for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in the flow or flows and / or block or blocks. Figure 1 The functions specified in the flow or flows and / or block or blocks.

[0230] These computer program instructions can also be stored in a computer-readable memory that can direct the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including instruction apparatus, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in the flow or flows and / or block or blocks. Figure 1 The functions specified in the flow or flows and / or block or blocks.

[0231] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable data processing devices, to generate computer-implemented processes, thus the instructions executed on the computer or other programmable data processing devices provide processes for implementing the functions specified in the flowchart Figure 1 one flow or a plurality of flows and / or the functions specified in the flowchart Figure 1 one flow or a plurality of flows and / or the functions specified in the flowchart

[0232] The above merely illustrates the embodiments of the present application, and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of the claims of the present application.

Claims

1. An environment perception system based on an autonomous driving vehicle, characterized in that: include: Visual processing module, radar processing module, joint processing module, autonomous driving decision module, data compression module, detection area division module and radar target tracking module; The visual processing module is configured to receive an image signal from the vehicle-mounted camera, perform a first image preprocessing on the image signal, and obtain image processing data generated by the image signal after the first image preprocessing; The radar processing module is used to perform a first data preprocessing on the environmental data collected by the radar sensor; The joint processing module is configured to perform a second data preprocessing on the outputs of the visual processing module and the radar processing module based on a multi-sensor fusion algorithm to generate environmental perception data; The autonomous driving decision module is used to combine the environmental perception data output by the joint processing module to generate control instructions to control the vehicle's response to target objects in different driving scenarios; The data compression module is configured to perform a second image preprocessing on the image processing data to obtain image compression data generated by the image processing data after the second image preprocessing; The detection area division module is configured to perform a third image preprocessing on the image compression data, and perform a second data preprocessing on the image compression data that has undergone the third image preprocessing, to determine the detection area; The radar target tracking module is used to monitor and classify objects in the detection area in real time based on the Kalman filter algorithm, and dynamically adjust the pre-defined tracking strategy according to the position, speed and motion trajectory parameters of the target object; Wherein, the first image preprocessing includes: adaptive adjustment of image resolution; The second image preprocessing includes: compressing the image data to remove noise and redundant information in the image; The third image preprocessing includes: dividing the image area; The first data processing includes: detecting distance, speed and position information of objects around the vehicle; The second data processing includes: using a multi-source data fusion algorithm to perform fusion processing on the data, converting the fused data into environmental perception data and forwarding it to the autonomous driving decision module; The predefined tracking strategy includes: based on the fact that the vehicle speed is proportional to the relative motion state of the target object during vehicle driving, according to the vehicle speed and acceleration parameters, dynamically adjusting the tracking accuracy and update frequency when the vehicle is in an accelerating or decelerating state.

2. The system according to claim 1, wherein: The visual processing module includes: a vehicle-mounted camera installed in the front and surrounding areas of the vehicle, and a collection submodule, an image processing submodule and an image data output submodule connected to the vehicle-mounted camera; The vehicle-mounted camera continuously collects image signals of the vehicle's surrounding environment at a specific frame rate; The acquisition submodule is used to acquire images through the vehicle-mounted camera and obtain image signals; The image processing submodule is used to process the high-resolution and low-resolution images in the image signal in parallel using an image processing algorithm to obtain high-resolution image processing data and low-resolution image processing data respectively; The image processing algorithm includes: a visual recognition method based on convolutional neural network (CNN) and a fast algorithm based on edge detection; The high-resolution image processing data includes: precise location information of distant objects and object category information; The low-resolution image processing data includes: edge features of close-range objects; The image data output submodule is used to output the image processing data obtained by the image processing submodule in the form of feature maps and position information. The output data consists of feature vectors and object detection frames in the format of ; Among them, x, y are the coordinates of the upper left corner of the object; w, h are the width and height of the object.

3. The system according to claim 2, characterized in that The image processing submodule includes: a first image processing unit for identifying long-distance obstacles and road condition information by extracting high-order features of the high-resolution image using the convolution layer of a deep convolutional network; after processing through several convolution and pooling layers, a set of deep features of the high-resolution image is obtained; The second image processing unit is used to process the close-range environment around the vehicle based on a fast edge detection algorithm for low-resolution images and perceive nearby vehicles, pedestrians and other moving objects.

4. The system according to claim 1, wherein: The radar processing module includes: Radar sensors are used to send electromagnetic waves and receive echo signals to detect the distance, speed, and position of objects in front of and around the vehicle; A radar signal input submodule is used to receive the echo signal from the radar sensor and process the echo signal according to the radar algorithm to generate preliminary radar scanning data; the radar scanning data includes the distance, angle and intensity information of the object; The visual data fusion submodule is used to receive the image processing data output by the image data output submodule and detect and identify objects by fusing it with the visual data; The radar algorithm application submodule is used to analyze obstacles in the detection area based on the radar algorithm and determine their position, speed and size.

5. The system according to claim 1, wherein: The joint processing module includes: The multi-source data fusion submodule is used to receive the output data of the radar processing module and the visual processing module and perform fusion processing; The environmental perception data generation submodule is used to convert the fused data into environmental perception data and output it to the autonomous driving decision module; wherein the environmental perception data includes the distance, angle, speed and position information of the object.

6. The system according to claim 1, wherein: The data compression module includes: A compression processing submodule, configured to receive high-resolution image data from the visual processing module and compress the data based on a specific compression algorithm; The compressed data output submodule is used to output the compressed data to the radar processing module; wherein the compressed data includes compressed image data in front of and around the vehicle.

7. The system according to claim 1, wherein: The detection area division module is composed of a background segmentation submodule and a region division submodule; The background segmentation submodule is used to binarize the image compression data, divide the pixels in the image into foreground pixels and background pixels, and after the binarization is completed, eliminate the noise in the binarized image through morphological opening operation, separate the background part and foreground objects in the binarized image, and obtain a relatively smooth and noise-free foreground image; The region division submodule is used to segment the image and distinguish different foreground regions according to the connection relationship between foreground pixels through a region labeling method based on connected domains; The method of binarizing the compressed image data and dividing the pixels in the image into foreground pixels and background pixels includes: assuming that the input grayscale image is , perform binarization operation according to the set threshold T, as follows: ; Where, For a binary image, the pixel value 1 represents the foreground pixel and the pixel value 0 represents the background pixel; Eliminating noise in a binary image by morphological opening operation includes: assuming the structural element is S, the formula of the opening operation is: ; Where, represents the corrosion operation, represents the expansion operation, is the binary image after opening operation.

8. The system according to claim 7, characterized in that The area division submodule includes: The search unit is used to search for all connected foreground pixel clusters in the binary image and mark each pixel cluster; assuming a pixel cluster is , all pixels inside it meet the connectivity condition, that is: ; Where d represents the distance between pixels and D is the connectivity threshold; Smoothing detection unit, used to effectively divide different areas of foreground pixels into several connected detection areas; using morphological closing operation to smooth the boundaries of the detection areas; The formula for the closing operation is: 。 9. The system according to claim 1, wherein: The radar target tracking module includes: The target recognition and tracking submodule is used to receive target object information from the detection area division module during the target recognition and tracking phase, and calculate the speed and position of the target object in real time based on the radar echo signal; The relative speed of the target object is calculated by the following formula: ; Where v is the relative velocity of the target object, is the frequency of the radar signal, c is the speed of light; is the measured Doppler frequency; is the frequency difference between the received echo signal and the transmitted signal; The spatial position of the target object is determined by the following formula: ; Where B is the bandwidth of the signal transmitted by the radar; d is the distance between the target and the radar; The radar target tracking submodule is used to track the trajectory of the target object through the Kalman filter and determine the estimated value of the object's motion state, as shown in the following formula: ; ; Where, is the state vector of the target, including position x, y, speed , Information, F is the state transfer matrix, is the process noise, Represents the measurement value at each moment, H is the observation matrix, To measure noise; The obstacle classification submodule is used to determine whether the target object is a moving object or a stationary object according to its motion state during the obstacle classification stage; for moving targets, it is classified according to its speed and motion characteristics.

10. An environment perception method based on an autonomous driving vehicle, characterized in that: The method comprises: receiving an image signal from a vehicle-mounted camera, performing a first image preprocessing on the image signal, and obtaining image processing data generated by the image signal after the first image preprocessing; Performing a first data preprocessing on the environmental data collected by the radar sensor; Based on a multi-sensor fusion algorithm, performing a second data preprocessing on the image processing data and a result of the first data preprocessing to generate environmental perception data; Combine environmental perception data to generate control commands to control the vehicle's response to target objects in different driving scenarios; performing a second image preprocessing on the image processing data to obtain image compression data generated by the image processing data after the second image preprocessing; performing a third image preprocessing on the image compression data, and performing a second data preprocessing on the image compression data that has undergone the third image preprocessing to determine a detection area; Based on the Kalman filter algorithm, objects in the detection area are monitored and classified in real time, and the pre-defined tracking strategy is dynamically adjusted according to the position, speed and motion trajectory parameters of the target object; Wherein, the first image preprocessing includes: adaptive adjustment of image resolution; The second image preprocessing includes: compressing the image data to remove noise and redundant information in the image; The third image preprocessing includes: dividing the image area; The first data processing includes: detecting distance, speed and position information of objects around the vehicle; The second data processing includes: using a multi-source data fusion algorithm to perform fusion processing on the data, converting the fused data into environmental perception data and forwarding it to the autonomous driving decision module; The predefined tracking strategy includes: based on the fact that the vehicle speed is proportional to the relative motion state of the target object during vehicle driving, according to the vehicle speed and acceleration parameters, dynamically adjusting the tracking accuracy and update frequency when the vehicle is in an accelerating or decelerating state.

Images