Vehicle blind area target identification method and system based on radar data

Abstract

The invention provides a vehicle blind area target identification method and system based on radar data. The method comprises the steps of synchronously collecting millimeter wave radar echo data and laser radar point cloud data at a vehicle blind area; performing compression and target information identification on the millimeter wave radar echo data to obtain first target data; sequentially performing grey-scale map generation, filtering and edge extraction on the laser radar point cloud data to obtain a target area grey-scale map; performing inverse mapping and attitude and trajectory estimation on the target area grey-scale map in sequence to obtain second target data; and fusing the first target data and the second target data to obtain target identification data, and realizing automobile blind area target detection based on high-speed maneuvering target robust detection and accurate point cloud rapid extraction.

Description

Technical Field

[0001] This invention belongs to the technical field of automotive driver assistance, specifically relating to a method and system for vehicle blind spot target recognition based on radar data. Background Technology

[0002] Blind spots are a major cause of traffic accidents. Existing blind spot monitoring systems mostly use a single sensor (such as millimeter-wave radar, ultrasonic radar, or a camera), which has the following shortcomings: Limitations of radar detection: Traditional moving target detection (MTD) and Keystone transform methods will experience distance movement and Doppler movement when the target is moving at high speed, which will lead to a decrease in detection sensitivity and make it easy to miss vehicles or pedestrians that are cutting in quickly.

[0003] Real-time issues in lidar point cloud processing: LiDAR can provide high-precision 3D point clouds, but the data volume is large, it is difficult to extract target point clouds in complex backgrounds, and the attitude changes of moving targets affect the extraction accuracy.

[0004] Single sensor information is insufficient: radar can provide distance and velocity information but lacks target outline, while lidar can provide fine shape but has a slow processing speed and is susceptible to background interference, making it difficult to simultaneously meet the requirements of real-time performance and accuracy. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a vehicle blind spot target recognition method and system based on radar data, which solves the technical problems in the prior art.

[0006] In a first aspect, the present invention provides the following technical solution: a method for vehicle blind spot target identification based on radar data, comprising: Simultaneously collect millimeter-wave radar echo data and lidar point cloud data in the blind spot of the vehicle; The millimeter-wave radar echo data is compressed and target information is identified to obtain the first target data; The point cloud data from the lidar is sequentially processed to generate a grayscale image, filter, and extract edges to obtain a grayscale image of the target area. The grayscale image of the target area is sequentially inversely mapped, and its attitude and trajectory are estimated to obtain the second target data. The first target data and the second target data are fused together to obtain target recognition data.

[0007] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention effectively compensates for distance movement and Doppler movement through three-dimensional motion parameter search and signal transformation, enabling robust detection even when the target is maneuvering at high speed. It is suitable for rapid target entry into blind zones. This invention rapidly removes background through projection dimensionality reduction and filtering, accurately matching the target point cloud. It has high processing efficiency and extraction accuracy superior to traditional methods. The radar detection results guide the lidar region of interest selection, significantly narrowing the search range. The lidar provides fine contour and attitude information, and the two complement each other, reducing false alarm rate and missed detection rate. This invention utilizes SVD decomposition to estimate target attitude changes, providing richer motion information for collision prediction.

[0008] Preferably, the step of compressing and identifying the millimeter-wave radar echo data to obtain the first target data specifically includes: The millimeter-wave radar echo data is compressed to obtain compressed data. : ; In the formula, They are fast time and slow time, respectively. For signal amplitude, For the Singer function, For signal bandwidth, For the instantaneous distance of the target, At the speed of light, It is an imaginary number. Wavelength; Several sets of search parameters are determined based on preset distance search intervals, preset radial velocity search intervals, and preset radial acceleration search intervals. , For the preset distance search interval, the first Distance data, For the preset radial velocity search interval, the first One radial velocity data point, For the preset radial acceleration search interval, the first One radial acceleration data point; Determine the target search distance for each set of search parameters. : ; Determine the target slow-time complex signal based on the target search distance. : ; In the formula, Compressed data corresponding to the target search distance; Based on the target slow-time complex signal Determine the primary target data.

[0009] Preferably, the method based on the target slow-time complex signal The steps to determine the first target data include: Determine the target slow-time complex signal First autocorrelation function With the second autocorrelation function : ; ; ; ; ; In the formula, The slow time after scaling transformation For Discrete Fourier Transform, These are the data signal function and the data conjugate signal function, respectively. , , For the first The amplitude center frequency and frequency modulation slope of a signal in the signal domain. As a scale factor, For time delay variables, It is a scaling constant; Based on the first autocorrelation function Second autocorrelation function Determine the representation of the windowed convolution domain : ; In the formula, These are the first and second center frequency window functions, respectively. These are the first autocorrelation function and the second autocorrelation function, respectively. The center frequency in the transform domain; The windowed convolutional domain representation along Perform a Fourier transform on the dimension to output the peak value in the domain; Traverse all search parameter groups and determine the corresponding domain peak value. When the domain peak value exceeds the preset peak value, it is determined that a target has been detected and the initial distance, radial velocity, and radial acceleration of the target are output to obtain the first target data.

[0010] Preferably, the step of sequentially generating a grayscale image, filtering, and extracting edges from the lidar point cloud data to obtain a grayscale image of the target region includes: The three-dimensional point cloud data of the lidar point cloud data Project onto the cylinder to obtain a projected scatter plot. : , ; In the formula, , These are the horizontal and vertical angular resolutions, respectively. By depth value or reflection intensity A scatter plot is determined for pixel values. The scatter plot is divided into grids at preset intervals. The pixel value of each grid is defined as the average value of all pixel values ​​within the grid to obtain a grayscale image. A two-dimensional fast Fourier transform is performed on the grayscale image to obtain a frequency domain image; Determine the mask : ; In the formula, For masking in The value at that location, The width of the frequency domain image. For column indexes; After multiplying the frequency domain image with the mask, an inverse Fourier transform is performed to obtain a filtered grayscale image. The edge image of the filtered grayscale image is extracted using a preset operator. The edge image is then binarized and connected component extraction is performed sequentially to obtain several connected components. The area and depth variation range of the connected components are determined, and the grayscale image of the target region is obtained by filtering based on the area and depth variation range of the connected components.

[0011] Preferably, the step of sequentially performing inverse mapping, pose and trajectory estimation on the grayscale image of the target region to obtain the second target data includes: Determine the mapping matrix : ; In the formula, A matrix constructed for the coordinates of the corner points of the scatter plot. The coordinate vector of the corner points of the grayscale image; The target region grayscale image is obtained according to the mapping matrix. Convert to target scatter plot : ; In the scatter plot, the nearest neighbor point is determined for each point in the target scatter plot and a point correspondence is established. Based on the point correspondence, the corresponding three-dimensional point cloud data is extracted from the lidar point cloud data to obtain the target point cloud set. The target point cloud is clustered using a preset clustering algorithm to obtain a filtered point cloud; Determine the maximum and minimum values ​​of the filtered point cloud in the X, Y, and Z directions, and determine the centroid coordinates of the filtered point cloud; Construct a standard cube vertex sequence based on the 3D point cloud data of the initial frame, and determine the vertex coordinates corresponding to the initial frame. And extract the vertex coordinates of the 3D point cloud data for each subsequent frame. ; Based on vertex coordinates and vertex coordinates Determine the initial vertex matrix and subsequent vertex matrices : ; ; Based on the initial vertex matrix and subsequent vertex matrices Determine the covariance matrix : ; Perform singular value decomposition on the covariance matrix and determine the rotation matrix based on the decomposition results. : ; ; In the formula, These are the first orthogonal matrix and the second orthogonal matrix in the decomposition result, respectively. To adjust the matrix; Based on the rotation matrix Determine the translation matrix : ; The target attitude angle is determined based on the rotation matrix, and the target trajectory is determined based on the translation matrix and the change of the centroid, so as to obtain the second target data.

[0012] Preferably, the step of fusing the first target data and the second target data to obtain target recognition data includes: The first target data and the second target data are fused using Kalman filtering to obtain target recognition data.

[0013] Secondly, the present invention provides the following technical solution: a vehicle blind spot target recognition system based on radar data, the system comprising: The acquisition module is used to simultaneously acquire millimeter-wave radar echo data and lidar point cloud data in the blind spot of the vehicle; The identification module is used to compress the millimeter-wave radar echo data and identify target information to obtain the first target data; The extraction module is used to sequentially generate grayscale images, filter, and extract edges from the lidar point cloud data to obtain grayscale images of the target area. The estimation module is used to sequentially perform inverse mapping, attitude and trajectory estimation on the grayscale image of the target area to obtain the second target data; The fusion module is used to fuse the first target data and the second target data to obtain target recognition data.

[0014] Preferably, the fusion module is specifically used for: The first target data and the second target data are fused using Kalman filtering to obtain target recognition data.

[0015] Thirdly, the present invention provides the following technical solution: a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the vehicle blind spot target recognition method based on radar data as described above.

[0016] Fourthly, the present invention provides the following technical solution: a storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the vehicle blind spot target recognition method based on radar data as described above. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a vehicle blind spot target recognition method based on radar data provided in Embodiment 1 of the present invention; Figure 2 This is a structural block diagram of a vehicle blind spot target recognition system based on radar data provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the hardware structure of a computer provided for another embodiment of the present invention.

[0019] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0021] Example 1 In Embodiment 1 of the present invention, as Figure 1 As shown, a vehicle blind spot target recognition method based on radar data includes: S1. Simultaneously collect millimeter-wave radar echo data and lidar point cloud data in the blind spot of the vehicle; Specifically, millimeter-wave radar echo data and lidar point cloud data can be obtained through millimeter-wave radar (installed on both sides of the rear bumper of the vehicle) and lidar (installed at the rear or side of the vehicle).

[0022] S2. Compress the millimeter-wave radar echo data and identify the target information to obtain the first target data; Step S2 includes: S21. Compress the millimeter-wave radar echo data to obtain compressed data. : ; In the formula, They are fast time and slow time, respectively. For signal amplitude, For the Singer function, For signal bandwidth, For the instantaneous distance of the target, At the speed of light, It is an imaginary number. Wavelength; Specifically, the purpose of compression is to achieve matched filtering.

[0023] S22. Determine several sets of search parameters based on preset distance search intervals, preset radial velocity search intervals, and preset radial acceleration search intervals. , For the preset distance search interval, the first Distance data, For the preset radial velocity search interval, the first One radial velocity data point, For the preset radial acceleration search interval, the first One radial acceleration data point; Specifically, in the actual search process, the search step size for the radial distance is... The search step size for the radial velocity is The search step size for radial acceleration is ,in, This refers to the time for coherent processing.

[0024] S23. Determine the target search distance for each group of search parameters. : .

[0025] S24. Determine the target slow-time complex signal based on the target search distance. : ; In the formula, Compressed data corresponding to the target search distance; Specifically, the target slow-time complex signal can be obtained by extracting from the compressed data obtained in the above steps.

[0026] S25. Based on the target slow-time complex signal Determine the primary target data; Step S25 includes: S251. Determine the target slow-time complex signal. First autocorrelation function With the second autocorrelation function : ; ; ; ; ; In the formula, The slow time after scaling transformation For Discrete Fourier Transform, These are the data signal function and the data conjugate signal function, respectively. , , For the first The amplitude center frequency and frequency modulation slope of a signal in the signal domain. As a scale factor, For time delay variables, This is a scaling constant.

[0027] S252, Based on the first autocorrelation function Second autocorrelation function Determine the representation of the windowed convolution domain : ; In the formula, These are the first and second center frequency window functions, respectively. These are the first autocorrelation function and the second autocorrelation function, respectively. The center frequency of the transform domain is given.

[0028] S253, Represent the windowed convolutional domain. along Perform a Fourier transform on the dimension to output the peak value in the domain; S254. Traverse all search parameter groups and determine the corresponding domain peak value. When the domain peak value exceeds the preset peak value, determine that a target has been detected and output the target's initial distance, radial velocity, and radial acceleration to obtain the first target data.

[0029] S3. The point cloud data of the lidar is sequentially processed to generate a grayscale image, filter and extract edges to obtain a grayscale image of the target area. Step S3 includes: S31, The three-dimensional point cloud data of the laser radar point cloud data Project onto the cylinder to obtain a projected scatter plot. : , ; In the formula, , These are the horizontal and vertical angular resolutions, respectively. S32, by depth value or reflection intensity A scatter plot is determined for pixel values. The scatter plot is divided into grids at preset intervals. The pixel value of each grid is defined as the average value of all pixel values ​​within the grid to obtain a grayscale image. Specifically, after defining the pixel values ​​of the grid, normalizing them to 0-255 will yield a grayscale image.

[0030] S33. Perform a two-dimensional fast Fourier transform on the grayscale image to obtain a frequency domain image; S34. Determine the mask. : ; In the formula, For masking in The value at that location, The width of the frequency domain image. For column indexes; S35. After multiplying the frequency domain image with the mask, perform an inverse Fourier transform to obtain a filtered grayscale image. S36. Use a preset operator to extract the edge image of the filtered grayscale image, perform image binarization and connected region extraction on the edge image in sequence to obtain several connected regions, determine the area and depth variation range of the connected regions, and filter according to the area and depth variation range of the connected regions to obtain the target region grayscale image. Specifically, the default operator here is the Sobel operator.

[0031] S4. Perform inverse mapping, attitude and trajectory estimation on the grayscale image of the target area in sequence to obtain the second target data; Step S4 includes: S41. Determine the mapping matrix : ; In the formula, A matrix constructed for the coordinates of the corner points of the scatter plot. The coordinate vector of the corner points of the grayscale image; Specifically, the scatter plot here is the same as the one in step S32.

[0032] S42. Based on the mapping matrix, the grayscale image of the target region is... Convert to target scatter plot : ; S43. In the scatter plot, determine the nearest neighbor point for each point in the target scatter plot and establish a point correspondence relationship. Based on the point correspondence relationship, extract the corresponding three-dimensional point cloud data from the lidar point cloud data to obtain the target point cloud set. S44. The target point cloud is clustered using a preset clustering algorithm to obtain a filtered point cloud; Specifically, the clustering algorithm used here is the K-means clustering algorithm.

[0033] S45. Determine the maximum and minimum values ​​of the filtered point cloud in the X, Y, and Z directions respectively, and determine the centroid coordinates of the filtered point cloud; S46. Construct a standard cube vertex sequence based on the 3D point cloud data of the initial frame, and determine the vertex coordinates corresponding to the initial frame. And extract the vertex coordinates of the 3D point cloud data for each subsequent frame. ; S47, Based on vertex coordinates and vertex coordinates Determine the initial vertex matrix and subsequent vertex matrices : ; ; S48, Based on the initial vertex matrix and subsequent vertex matrices Determine the covariance matrix : ; S49. Perform singular value decomposition on the covariance matrix and determine the rotation matrix based on the decomposition results. : ; ; In the formula, These are the first orthogonal matrix and the second orthogonal matrix in the decomposition result, respectively. To adjust the matrix; S410, Based on the rotation matrix Determine the translation matrix : ; S411. Determine the target attitude angle based on the rotation matrix, and determine the target trajectory based on the translation matrix and the change of the centroid, so as to obtain the second target data.

[0034] S5. The first target data and the second target data are fused together to obtain target recognition data; Step S5 includes: The first target data and the second target data are fused using Kalman filtering to obtain target recognition data.

[0035] The vehicle blind spot target recognition method based on radar data provided in Embodiment 1 of this invention effectively compensates for distance movement and Doppler movement through three-dimensional motion parameter search and signal transformation, enabling robust detection even when the target is maneuvering at high speed. It is suitable for rapid target entry into blind spots. This invention uses projection dimensionality reduction and filtering to quickly remove the background and accurately match the target point cloud, resulting in high processing efficiency and extraction accuracy superior to traditional methods. The radar detection results guide the lidar region of interest screening, significantly narrowing the search range. The lidar provides fine contour and attitude information, and the two complement each other, reducing the false alarm rate and missed detection rate. This invention utilizes SVD decomposition to estimate target attitude changes, providing richer motion information for collision prediction.

[0036] Example 2 like Figure 2 As shown, in Embodiment 2 of the present invention, a vehicle blind spot target recognition system based on radar data is provided, the system comprising: Acquisition module 1 is used to simultaneously acquire millimeter-wave radar echo data and lidar point cloud data in the blind spot of the vehicle; The identification module 2 is used to compress the millimeter-wave radar echo data and identify target information to obtain the first target data; Extraction module 3 is used to sequentially generate grayscale images, filter, and extract edges from the lidar point cloud data to obtain grayscale images of the target area. Estimation module 4 is used to sequentially perform inverse mapping, attitude and trajectory estimation on the grayscale image of the target area to obtain the second target data; The fusion module 5 is used to fuse the first target data and the second target data to obtain target recognition data.

[0037] The identification module 2 is used for: The millimeter-wave radar echo data is compressed to obtain compressed data. : ; In the formula, They are fast time and slow time, respectively. For signal amplitude, For the Singer function, For signal bandwidth, For the instantaneous distance of the target, At the speed of light, It is an imaginary number. Wavelength; Several sets of search parameters are determined based on preset distance search intervals, preset radial velocity search intervals, and preset radial acceleration search intervals. , For the preset distance search interval, the first Distance data, For the preset radial velocity search interval, the first One radial velocity data point, For the preset radial acceleration search interval, the first One radial acceleration data point; Determine the target search distance for each set of search parameters. : ; Determine the target slow-time complex signal based on the target search distance. : ; In the formula, Compressed data corresponding to the target search distance; Based on the target slow-time complex signal Determine the primary target data.

[0038] The identification module 2 is also used for: Determine the target slow-time complex signal First autocorrelation function With the second autocorrelation function : ; ; ; ; ; In the formula, The slow time after scaling transformation For Discrete Fourier Transform, These are the data signal function and the data conjugate signal function, respectively. , , For the first The amplitude center frequency and frequency modulation slope of a signal in the signal domain. As a scale factor, For time delay variables, It is a scaling constant; Based on the first autocorrelation function Second autocorrelation function Determine the representation of the windowed convolution domain : ; In the formula, These are the first and second center frequency window functions, respectively. These are the first autocorrelation function and the second autocorrelation function, respectively. The center frequency in the transform domain; The windowed convolutional domain representation along Perform a Fourier transform on the dimension to output the peak value in the domain; Traverse all search parameter groups and determine the corresponding domain peak value. When the domain peak value exceeds the preset peak value, it is determined that a target has been detected and the initial distance, radial velocity, and radial acceleration of the target are output to obtain the first target data.

[0039] The extraction module 3 is used for: The three-dimensional point cloud data of the lidar point cloud data Project onto the cylinder to obtain a projected scatter plot. : , ; In the formula, , These are the horizontal and vertical angular resolutions, respectively. By depth value or reflection intensity A scatter plot is determined for pixel values. The scatter plot is divided into grids at preset intervals. The pixel value of each grid is defined as the average value of all pixel values ​​within the grid to obtain a grayscale image. A two-dimensional fast Fourier transform is performed on the grayscale image to obtain a frequency domain image; Determine the mask : ; In the formula, For masking in The value at that location, The width of the frequency domain image. For column indexes; After multiplying the frequency domain image with the mask, an inverse Fourier transform is performed to obtain a filtered grayscale image. The edge image of the filtered grayscale image is extracted using a preset operator. The edge image is then binarized and connected component extraction is performed sequentially to obtain several connected components. The area and depth variation range of the connected components are determined, and the grayscale image of the target region is obtained by filtering based on the area and depth variation range of the connected components.

[0040] The estimation module 4 is used for: Determine the mapping matrix : ; In the formula, A matrix constructed for the coordinates of the corner points of the scatter plot. The coordinate vector of the corner points of the grayscale image; The target region grayscale image is obtained according to the mapping matrix. Convert to target scatter plot : ; In the scatter plot, the nearest neighbor point is determined for each point in the target scatter plot and a point correspondence is established. Based on the point correspondence, the corresponding three-dimensional point cloud data is extracted from the lidar point cloud data to obtain the target point cloud set. The target point cloud is clustered using a preset clustering algorithm to obtain a filtered point cloud; Determine the maximum and minimum values ​​of the filtered point cloud in the X, Y, and Z directions, and determine the centroid coordinates of the filtered point cloud; Construct a standard cube vertex sequence based on the 3D point cloud data of the initial frame, and determine the vertex coordinates corresponding to the initial frame. And extract the vertex coordinates of the 3D point cloud data for each subsequent frame. ; Based on vertex coordinates and vertex coordinates Determine the initial vertex matrix and subsequent vertex matrices : ; ; Based on the initial vertex matrix and subsequent vertex matrices Determine the covariance matrix : ; Perform singular value decomposition on the covariance matrix and determine the rotation matrix based on the decomposition results. : ; ; In the formula, These are the first orthogonal matrix and the second orthogonal matrix in the decomposition result, respectively. To adjust the matrix; Based on the rotation matrix Determine the translation matrix : ; The target attitude angle is determined based on the rotation matrix, and the target trajectory is determined based on the translation matrix and the change of the centroid, so as to obtain the second target data.

[0041] The fusion module 5 is used for: The first target data and the second target data are fused using Kalman filtering to obtain target recognition data.

[0042] In other embodiments of the present invention, the present invention provides the following technical solution: a computer, including a memory 102, a processor 101, and a computer program stored in the memory 102 and executable on the processor 101, wherein the processor 101 executes the computer program to implement the vehicle blind spot target recognition method based on radar data as described above.

[0043] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0044] The memory 102 may include a large-capacity memory for data or instructions. For example, and not limitingly, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 102 may include removable or non-removable (or fixed) media. Where appropriate, the memory 102 may be internal or external to a data processing device. In a particular embodiment, the memory 102 is non-volatile memory. In a particular embodiment, the memory 102 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random Access Memory (FPMDRAM), Extended Data Out Dynamic Random Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0045] The memory 102 can be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101.

[0046] The processor 101 reads and executes the computer program instructions stored in the memory 102 to implement the above-mentioned vehicle blind spot target recognition method based on radar data.

[0047] In some embodiments, the computer may further include a communication interface 103 and a bus 100. For example, Figure 3 As shown, the processor 101, memory 102, and communication interface 103 are connected through bus 100 and complete communication with each other.

[0048] The communication interface 103 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 103 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0049] Bus 100 includes hardware, software, or both, that couples components of a computer device together. Bus 100 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 100 may include one or more buses. Although specific buses are described and illustrated in the embodiments of the present invention, the present invention is contemplated by any suitable bus or interconnect.

[0050] The computer can execute the vehicle blind spot target recognition method based on radar data of the present invention based on the vehicle blind spot target recognition system that has acquired radar data, thereby realizing vehicle blind spot target recognition based on radar data.

[0051] In some further embodiments of the present invention, in conjunction with the above-described vehicle blind spot target recognition method based on radar data, the present invention provides the following technical solution: a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the above-described vehicle blind spot target recognition method based on radar data.

[0052] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0053] More specific examples of readable media (a non-exhaustive list) include: electrical connections (electronic devices) with one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0054] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0055] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0056] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for vehicle blind spot target identification based on radar data, characterized in that, include: Simultaneously collect millimeter-wave radar echo data and lidar point cloud data in the blind spot of the vehicle; The millimeter-wave radar echo data is compressed and target information is identified to obtain the first target data; The point cloud data from the lidar is sequentially processed to generate a grayscale image, filter, and extract edges to obtain a grayscale image of the target area. The grayscale image of the target area is sequentially inversely mapped, and its attitude and trajectory are estimated to obtain the second target data. The first target data and the second target data are fused together to obtain target recognition data.

2. The vehicle blind spot target recognition method based on radar data according to claim 1, characterized in that, The step of compressing the millimeter-wave radar echo data and identifying target information to obtain the first target data specifically includes: The millimeter-wave radar echo data is compressed to obtain compressed data. : ； In the formula, They are fast time and slow time, respectively. For signal amplitude, For the Singer function, For signal bandwidth, For the instantaneous distance of the target, At the speed of light, It is an imaginary number. Wavelength; Several sets of search parameters are determined based on preset distance search intervals, preset radial velocity search intervals, and preset radial acceleration search intervals. , For the preset distance search interval, the first Distance data, For the preset radial velocity search interval, the first One radial velocity data, For the preset radial acceleration search interval, the first One radial acceleration data point; Determine the target search distance for each set of search parameters. : ； Determine the target slow-time complex signal based on the target search distance. : ； In the formula, Compressed data corresponding to the target search distance; Based on the target slow-time complex signal Determine the primary target data.

3. The vehicle blind spot target recognition method based on radar data according to claim 2, characterized in that, The target slow-time complex signal The steps to determine the first target data include: Determine the target slow-time complex signal First autocorrelation function With the second autocorrelation function : ； ； ； ； ； In the formula, The slow time after scaling transformation For Discrete Fourier Transform, These are the data signal function and the data conjugate signal function, respectively. , , For the first The amplitude center frequency and frequency modulation slope of a signal in the signal domain. As a scale factor, For time delay variables, It is a scaling constant; Based on the first autocorrelation function Second autocorrelation function Determine the windowed convolution domain representation : ； In the formula, These are the first and second center frequency window functions, respectively. These are the first autocorrelation function and the second autocorrelation function, respectively. The center frequency in the transform domain; The windowed convolutional domain representation along Perform a Fourier transform on the dimension to output the peak value in the domain; Traverse all search parameter groups and determine the corresponding domain peak value. When the domain peak value exceeds the preset peak value, it is determined that a target has been detected and the initial distance, radial velocity, and radial acceleration of the target are output to obtain the first target data.

4. The vehicle blind spot target recognition method based on radar data according to claim 1, characterized in that, The steps of sequentially generating a grayscale image, filtering, and extracting edges from the lidar point cloud data to obtain a grayscale image of the target region include: The three-dimensional point cloud data of the lidar point cloud data Project onto the cylinder to obtain a projected scatter plot. : ， ； In the formula, , These are the horizontal and vertical angular resolutions, respectively. By depth value or reflection intensity A scatter plot is determined for pixel values. The scatter plot is divided into grids at preset intervals. The pixel value of each grid is defined as the average value of all pixel values ​​within the grid to obtain a grayscale image. A two-dimensional fast Fourier transform is performed on the grayscale image to obtain a frequency domain image; Determine the mask : ； In the formula, For masking in The value at that location, The width of the frequency domain image. For column indexes; After multiplying the frequency domain image with the mask, an inverse Fourier transform is performed to obtain a filtered grayscale image. The edge image of the filtered grayscale image is extracted using a preset operator. The edge image is then binarized and connected component extraction is performed sequentially to obtain several connected components. The area and depth variation range of the connected components are determined, and the grayscale image of the target region is obtained by filtering based on the area and depth variation range of the connected components.

5. The vehicle blind spot target recognition method based on radar data according to claim 1, characterized in that, The step of sequentially performing inverse mapping, pose and trajectory estimation on the grayscale image of the target region to obtain the second target data includes: Determine the mapping matrix : ； In the formula, A matrix constructed for the coordinates of the corner points of the scatter plot. The coordinate vector of the corner points of the grayscale image; The target region grayscale image is obtained according to the mapping matrix. Convert to target scatter plot : ； In the scatter plot, the nearest neighbor point is determined for each point in the target scatter plot and a point correspondence is established. Based on the point correspondence, the corresponding three-dimensional point cloud data is extracted from the lidar point cloud data to obtain the target point cloud set. The target point cloud is clustered using a preset clustering algorithm to obtain a filtered point cloud; Determine the maximum and minimum values ​​of the filtered point cloud in the X, Y, and Z directions, and determine the centroid coordinates of the filtered point cloud; Construct a standard cube vertex sequence based on the 3D point cloud data of the initial frame, and determine the vertex coordinates corresponding to the initial frame. And extract the vertex coordinates of the 3D point cloud data for each subsequent frame. ; Based on vertex coordinates and vertex coordinates Determine the initial vertex matrix and subsequent vertex matrices : ； ； Based on the initial vertex matrix and subsequent vertex matrices Determine the covariance matrix : ； Perform singular value decomposition on the covariance matrix and determine the rotation matrix based on the decomposition results. : ； ； In the formula, These are the first orthogonal matrix and the second orthogonal matrix in the decomposition result, respectively. To adjust the matrix; Based on the rotation matrix Determine the translation matrix : ； The target attitude angle is determined based on the rotation matrix, and the target trajectory is determined based on the translation matrix and the change of the centroid, so as to obtain the second target data.

6. The vehicle blind spot target recognition method based on radar data according to claim 1, characterized in that, The step of fusing the first target data and the second target data to obtain target recognition data includes: The first target data and the second target data are fused using Kalman filtering to obtain target recognition data.

7. A vehicle blind spot target recognition system based on radar data, characterized in that, The system includes: The acquisition module is used to simultaneously acquire millimeter-wave radar echo data and lidar point cloud data in the blind spot of the vehicle; The identification module is used to compress the millimeter-wave radar echo data and identify target information to obtain the first target data; The extraction module is used to sequentially generate grayscale images, filter, and extract edges from the lidar point cloud data to obtain grayscale images of the target area. The estimation module is used to sequentially perform inverse mapping, attitude and trajectory estimation on the grayscale image of the target area to obtain the second target data; The fusion module is used to fuse the first target data and the second target data to obtain target recognition data.

8. The vehicle blind spot target recognition system based on radar data according to claim 7, characterized in that, The fusion module is specifically used for: The first target data and the second target data are fused using Kalman filtering to obtain target recognition data.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the vehicle blind spot target recognition method based on radar data as described in any one of claims 1 to 6.

10. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the vehicle blind spot target recognition method based on radar data as described in any one of claims 1 to 6.

Images