Automobile millimeter wave radar signal processing method and system
Through triangular pulse sampling and fast Fourier transform technology, combined with multi-dimensional filtering and constant false alarm rate detection, the problems of insufficient computing power and low distance resolution in traditional radar signal processing are solved, and efficient point cloud output and target are achieved Classification improves radar detection performance.
Patent Information
- Application Number
- CN202411068249.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-12
AI Technical Summary
Traditional radar signal processing has problems such as insufficient computing power, large loss of distance/speed window function, and inability to fully improve radar distance resolution, especially when the number of MIMO channels increases, the distance measurement and speed measurement performance is limited.
Triangular pulse sampling generates the distance-Doppler data matrix, which is converted into radar data cubes through distance-dimensional fast Fourier transform and data compression, and is decompressed in batches based on the distance cell distribution, and performs multi-dimensional filtering, phase compensation and constant False alarm rate detection, output point cloud list; combine secondary clustering association and management of vehicle tracks, output target classification results.
It improves the signal-to-noise ratio and distance resolution of radar signal processing, improves the accuracy and detection performance of point cloud output, enhances the angle measurement accuracy of large field of view angles, and reduces false alarm rate.
Smart Images

Figure CN118938160A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of radar signal processing, and in particular to a method and system for processing automobile millimeter-wave radar signals. Background Art
[0002] In recent years, with the gradual development of the new energy vehicle industry and the orderly implementation of pilot projects in various regions, my country's autonomous driving field has developed well, and its technology research and development and application are at the forefront of the world. From a global perspective, unmanned driving / intelligent driving / autonomous driving is the most cutting-edge technology in the current automotive industry and is also one of the main application scenarios of artificial intelligence.
[0003] Starting from 2021, the emergence of 4D imaging millimeter wave radar technology will not only change the field of autonomous driving, but also drive the value of the entire automotive industry and promote the construction of intelligent transportation and smart cities. In addition to the distance measurement, speed measurement, and azimuth (3D) measurement of traditional radars, 4D imaging millimeter wave radars can also measure pitch angles, adding a 1D measurement dimension. 4D imaging millimeter wave radars generally adopt a multi-chip cascade solution, and the number of transceiver channels ranges from 48 to 2000, far exceeding the 12-16 of traditional 3D radars. It can achieve accurate perception and identification of various traffic participants such as vehicles, pedestrians, bicycles, and motorcycles, thereby improving traffic safety and efficiency, reducing traffic accidents and congestion, reducing energy consumption and emissions, and improving travel comfort and convenience. 4D imaging millimeter wave radars can also be combined with technologies such as Internet of Vehicles, cloud computing, and big data to achieve real-time collection, analysis, and sharing of traffic information, thereby providing strong support for the planning, management, and services of intelligent transportation and smart cities. In addition, 4D imaging millimeter-wave radar can also be applied to other fields, such as security, medical care, industry, agriculture, etc., showing broad market prospects and social value.
[0004] The signal processing flow has a decisive influence on the performance of 4D imaging millimeter wave radar. The signal processing flow reflects the essence of the radar system. The 4D imaging radar signal processing module receives the ADC data input by the RF front end, outputs the point cloud data to the external interface, and can also output the point cloud to the data processing module to form a track target and then output it to the external interface.
[0005] The signal processing of traditional radar has the following disadvantages: insufficient computing power usually means that coherent accumulation between channels is not performed; the window function loss after windowing in the distance / speed dimension is relatively large; it is usually a TD MIMO system, so the ranging and speed measurement performance is limited when the number of MIMO channels (wireless communication channels) increases; the waveform uses conventional FMCW (frequency modulated continuous wave) rather than frequency hopping FMCW, so the radar distance resolution cannot be fully improved, etc.
[0006] Therefore, how to provide a method and system for processing automotive millimeter-wave radar signals is a problem that needs to be solved urgently. Summary of the invention
[0007] The embodiments of the present invention provide a method and system for processing automotive millimeter-wave radar signals to solve the above-mentioned technical problems existing in the prior art.
[0008] In order to have a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended to be a general review, nor is it intended to identify key / important components or to delineate the scope of protection of these embodiments. Its only purpose is to present some concepts in a simple form as a preface to the detailed description that follows.
[0009] According to a first aspect of an embodiment of the present invention, a method for processing automotive millimeter-wave radar signals is provided.
[0010] In one embodiment, the automotive millimeter-wave radar signal processing method includes:
[0011] The range-Doppler data matrix is generated by triangular pulse sampling and converted into radar data cube based on range-dimensional fast Fourier transform and data compression;
[0012] Based on the range cell distribution, radar data cubes are selected in batches for decompression, and point cloud lists are output through multi-dimensional filtering, phase compensation and constant false alarm rate detection;
[0013] The radar point cloud list is synchronized with the vehicle body information, and based on secondary clustering, the vehicle track is associated and managed to output the target classification results.
[0014] In one embodiment, the generating of the range-Doppler data matrix by triangular pulse sampling and converting it into a radar data cube based on range-dimensional fast Fourier transform and data compression comprises:
[0015] Acquire the triangular pulses in the wireless communication channel, sample the sampling points of all triangular pulses in one frame, and obtain the range-Doppler data matrix;
[0016] Perform a distance dimension fast Fourier transform on each column of the range-Doppler data matrix to calculate the distribution of the detected target in each range cell;
[0017] The calculation results of the range-dimensional fast Fourier transform of the triangular pulse are compressed to obtain the radar data cube.
[0018] In one embodiment, based on the distance cell distribution, radar data cubes are selected in batches for decompression, and through multi-dimensional filtering, phase compensation and constant false alarm rate detection, the output point cloud list includes:
[0019] Select radar data cells of a number of continuous range cells in batches and decompress them;
[0020] Perform a velocity-dimensional fast Fourier transform on the decompressed range-Doppler matrix, apply spatial apodization filtering to the calculation result, and generate a noise spectrum for each range cell based on histogram statistics;
[0021] Set a threshold value, remove target points below the threshold value in the range-Doppler matrix, retain target points above the threshold value as coarsely selected targets, and proceed to the next step;
[0022] Decode the transmission channel sequence and perform phase compensation on the wireless communication channel;
[0023] Based on the angle-dimensional fast Fourier transform, the intermediate angle is calculated and constant false alarm rate detection is performed;
[0024] Constructing an elevation array, and based on the elevation array, calculating the elevation angle and azimuth angle corresponding to the intermediate angle;
[0025] Based on the optimization of angle measurement accuracy and the elimination of velocity ambiguity, a point cloud list is output;
[0026] The radar data cells of the next batch of several continuous distance cells are selected again, decompressed, and the distance, speed and angle calculations are repeated until all distance cells are traversed.
[0027] In one embodiment, performing a velocity-dimensional fast Fourier transform on the decompressed range-Doppler matrix, applying a spatial apodization filter to the calculation result, and generating a noise spectrum for each range cell based on histogram statistics includes:
[0028] Perform a velocity-dimensional fast Fourier transform on each row of data in the range-Doppler matrix to calculate the distribution of the detected target on the velocity cell;
[0029] The velocity dimension sidelobes of the velocity dimension calculation results are suppressed by using spatial apodization filtering;
[0030] The histogram is used to count all the speed values in each distance cell, the noise level of the distance cell is estimated, and the noise spectrum of the entire distance cell is obtained;
[0031] The range dimension apodization filtering is used to suppress the range dimension sidelobes of the velocity dimension calculation results.
[0032] In one embodiment, the phase compensation includes: initial phase calibration and one-degree calibration of the azimuth angle with a large field of view.
[0033] In one embodiment, the calculation of the intermediate angle based on the angle-dimensional fast Fourier transform and the constant false alarm rate detection include:
[0034] Set the sub-array and arrangement vector, calculate the middle angle of the detected target by performing angle-dimensional fast Fourier transform on the range-Doppler matrix, and perform inter-channel coherence accumulation;
[0035] Perform constant false alarm rate detection in the speed dimension and angle dimension to perform precise inspection on the detection target;
[0036] Through the super-resolution measurement method, super-resolution processing is performed based on the intermediate angle.
[0037] In one embodiment, constructing the elevation array and calculating the elevation angle and azimuth angle corresponding to the intermediate angle based on the elevation array includes:
[0038] Multiply the sub-arrays at each elevation height by the steering vector of the middle angle and align them to form an elevation array;
[0039] Perform fast Fourier transform on the pitch array to obtain all pitch angles corresponding to the intermediate angle;
[0040] Based on the calculated pitch angle, the azimuth angle corresponding to the intermediate angle is decoupled.
[0041] In one embodiment, based on the optimization of angle measurement accuracy and elimination of velocity ambiguity, the output point cloud list includes:
[0042] Based on the distance, speed and angle calculation results, secondary interpolation processing is performed;
[0043] Based on the pattern beamforming technology, the angle measurement accuracy of a large field of view is calibrated;
[0044] Through frequency hopping decoupling and dual-frequency wave matching, velocity ambiguity is eliminated and a point cloud list is output.
[0045] In one embodiment, synchronizing the radar point cloud list with the vehicle body information, associating and managing the vehicle track based on secondary clustering, and outputting the target classification result includes:
[0046] Synchronize vehicle body information to radar and correct radar measured values;
[0047] Through preprocessing, outliers and false points are removed, and point cloud classification is performed based on vehicle body information;
[0048] Set a clustering radius, and remove detection targets that do not meet the clustering requirements through clustering;
[0049] Based on the detected target of the previous frame, the state of the current frame is predicted, and the point traces are associated with the track. Based on the clustering algorithm, the point traces are clustered. If there is a track in the clustering result, a new track will not be started.
[0050] Manage the state transition and judgment of the track, and output the target type based on the target classification results.
[0051] According to a second aspect of an embodiment of the present invention, a system for processing automotive millimeter-wave radar signals is provided.
[0052] In one embodiment, the automotive millimeter wave radar signal processing system includes:
[0053] The first-level signal processing module is used to generate a range-Doppler data matrix through triangular pulse sampling, and convert it into a radar data cube based on range-dimensional fast Fourier transform and data compression;
[0054] The secondary signal processing module is used to select radar data cubes in batches for decompression based on the distance cell distribution, and output the point cloud list through multi-dimensional filtering, phase compensation and constant false alarm rate detection;
[0055] The data processing module is used to synchronize the radar point cloud list with the vehicle body information, perform preprocessing and point cloud classification, and associate and manage vehicle tracks based on secondary clustering to output target classification results.
[0056] According to a third aspect of an embodiment of the present invention, a computer device is provided.
[0057] In some embodiments, the computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.
[0058] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention aims at the deficiencies of traditional automotive radar signal processing and data processing, combines the latest system and technology of automotive radar signal processing, and proposes a new processing method, which fully utilizes the computing power resources of existing automotive millimeter-wave radar processor chips to quickly and accurately complete the output of imaging radar point clouds and tracks. Specifically, it can be summarized as follows:
[0059] By adopting the DDMAMIMO system, the signal-to-noise ratio is increased by N times compared with traditional automotive radars; the frequency-hopping FMCW waveform is used to greatly improve the distance resolution without reducing the maximum ranging range; the DML super-resolution algorithm is used, and the angular resolution is expected to be increased by 1 times; the SVA window function is used to avoid the signal-to-noise ratio and resolution loss of traditional Hanning / Hamming windows; at the same time, the DA-CFAR module using the histogram statistics method is used to obtain the noise spectrum of the entire distance segment, and the cfar performance is better than the traditional OS-CFAR and CA-CFAR;
[0060] By compressing the distance dimension fast Fourier transform result by 2 times, the radar data volume / signal-to-noise ratio is improved; by combining the coarse screening Coarse RD CFAR with the fine screening DA CFAR, more point clouds are released and the radar detection performance is improved at the same time; the angle measurement module combines the advantages of the fast Fourier transform algorithm and the combined mode, and the angle measurement is faster and more accurate; in view of the fact that the antenna phase characteristics deteriorate at a large field of view angle (such as 45°-75°), the present invention re-measures the targets in this angle range using a degree-calibrated DBF (directional pattern beamforming) angle measurement method to improve the accuracy of large-angle angle measurement.
[0061] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0063] Figure 1 is a flow chart of a method for processing automotive millimeter-wave radar signals according to an exemplary embodiment;
[0064] Figure 2 is a system principle block diagram of a vehicle millimeter wave radar signal processing system according to an exemplary embodiment;
[0065] Figure 3 is a flowchart of conventional radar signal processing and data processing according to an exemplary embodiment;
[0066] Figure 4 is a flow chart of 4D imaging radar signal processing and data processing according to an exemplary embodiment of the present invention;
[0067] Figure 5 is a frequency hopping FMCW waveform diagram of each frame of a MIMO channel according to an exemplary embodiment;
[0068] Figure 6 is a range-Doppler data matrix of a MIMO channel / element according to an exemplary embodiment;
[0069] Figure 7 is a range-Doppler matrix of a MIMO channel / unit after decompression of 10 range bins according to an exemplary embodiment;
[0070] Figure 8 is a schematic diagram of histogram statistics according to an exemplary embodiment;
[0071] Fig. 9It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 7th and 8th sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 0th-31st speed bin);
[0072] Fig.10 It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 1st and 8th sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 32nd-63rd speed bin);
[0073] Fig.11 It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 1st and 2nd sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 64th-95th speed bin);
[0074] Fig.12 It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 2nd and 3rd sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 96th-127th speed bin);
[0075] Fig.13 It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 3rd and 4th sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 128th-159th speed bin);
[0076] Fig.14 It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 4th and 5th sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 160th-191st speed bin);
[0077] Fig.15 It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 5th and 6th sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 192th-223rd speed bin);
[0078] Fig.16 It is a distribution diagram of each TX channel on the doppler frequency band when the empty sub-band occupies the 6th and 7th sub-bands according to an exemplary embodiment (corresponding to the actual target speed falling in the 224th-255th speed bin);
[0079] Fig.17 According to an exemplary embodiment, the data of each MIMO channel is taken and the data of the same distance unit and velocity unit cell (black) on the range-Doppler matrix are taken to perform angle dimension FFT;
[0080] Fig.18is a schematic diagram of interpolation according to an exemplary embodiment;
[0081] Fig.19 It is a schematic diagram showing the structure of a computer device according to an exemplary embodiment. DETAILED DESCRIPTION
[0082] The following description and accompanying drawings fully illustrate the specific embodiments of this article so that those skilled in the art can practice them. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. The scope of the embodiments of this article includes the entire scope of the claims, as well as all available equivalents of the claims. Herein, the terms "first", "second", etc. are only used to distinguish one element from another, without requiring or implying any actual relationship or order between these elements. In fact, the first element can also be called the second element, and vice versa. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the structure, device or equipment including a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also include elements inherent to such structure, device or equipment. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the structure, device or equipment including the elements. Each embodiment is described in a progressive manner herein, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the embodiments can be referred to each other.
[0083] The terms "longitudinal", "lateral", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. in this document indicate the orientation or position relationship based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing this document and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In the description of this document, unless otherwise specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a mechanical connection or an electrical connection, it can also be the internal communication of two elements, it can be a direct connection, or it can be an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.
[0084] As used herein, unless otherwise specified, the term "plurality" means two or more than two.
[0085] In this document, the character " / " indicates that the preceding and following objects are in an "or" relationship. For example, A / B means: A or B.
[0086] In this article, the term "and / or" is a description of the association relationship between objects, indicating that three relationships may exist. For example, A and / or B means: A or B, or, A and B.
[0087] It should be understood that, although the various steps in the flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the figure may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
[0088] Each module in the device or system of the present application can be implemented in whole or in part by software, hardware, or a combination thereof. The above modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above modules.
[0089] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0090] Figure 1 An embodiment of a method for processing automotive millimeter-wave radar signals according to the present invention is shown.
[0091] In this optional embodiment, the automotive millimeter wave radar signal processing method includes:
[0092] Step S101, generating a range-Doppler data matrix by triangular pulse sampling, and converting it into a radar data cube based on range-dimensional fast Fourier transform and data compression;
[0093] Step S103: based on the distance cell distribution, radar data cubes are selected in batches for decompression, and a point cloud list is output through multi-dimensional filtering, phase compensation and constant false alarm rate detection;
[0094] Step S105: synchronize the radar point cloud list with the vehicle body information, associate and manage the vehicle track based on secondary clustering, and output the target classification result.
[0095] In this optional embodiment, when the range-Doppler data matrix is generated by triangular pulse sampling and converted into a radar data cube based on range-dimensional fast Fourier transform and data compression, the triangular pulses in the wireless communication channel can be obtained, and the sampling points of all triangular pulses in one frame are sampled to obtain the range-Doppler data matrix; the range-dimensional fast Fourier transform is performed on each column of the range-Doppler data matrix to calculate the distribution of the detection target in each range cell; the calculation result of the range-dimensional fast Fourier transform of the triangular pulse is compressed to obtain the radar data cube.
[0096] In this optional embodiment, when the radar data cubes are selected in batches for decompression based on the distribution of the distance cells, and the radar data cells of a number of continuous distance cells are selected in batches for decompression through multi-dimensional filtering, phase compensation and constant false alarm rate detection, when the point cloud list is output; the decompressed range-Doppler matrix is subjected to a velocity-dimensional fast Fourier transform, a spatial apodization filter is applied to the calculation result, and a noise spectrum of each distance cell is generated based on histogram statistics; a threshold value is set to eliminate the target points below the threshold value in the range-Doppler matrix, and retain Target points above the threshold value are taken as coarsely selected targets and enter the next step; the transmission channel sequence is solved and the phase compensation is performed on the wireless communication channel; the intermediate angle is calculated based on the angle-dimensional fast Fourier transform, and constant false alarm rate detection is performed; a pitch array is constructed, and the pitch angle and azimuth corresponding to the intermediate angle are calculated based on the pitch array; a point cloud list is output based on the optimization of angle measurement accuracy and the elimination of velocity ambiguity; the radar data cells of the next batch of consecutive distance cells are selected again, decompressed, and the distance, velocity and angle calculations are repeated until all distance cells are traversed.
[0097] In this optional embodiment, when performing a speed-dimensional fast Fourier transform on the decompressed range-Doppler matrix, applying a spatial apodization filter to the calculation result, and generating a noise spectrum for each range cell segment based on histogram statistics, a speed-dimensional fast Fourier transform can be performed on each row of data in the range-Doppler matrix to calculate the distribution of the detection target on the speed cell segment; the speed-dimensional side lobes of the speed-dimensional calculation result are suppressed by using the spatial apodization filter; all speed values on each range cell segment are counted by using the histogram to estimate the noise level of the range cell segment and obtain the noise spectrum for the entire range cell segment; and the range-dimensional apodization filter is used to suppress the range-dimensional side lobes of the speed-dimensional calculation result.
[0098] In this optional embodiment, the phase compensation includes: initial phase calibration and one-degree calibration of the azimuth angle with a large field of view.
[0099] In this optional embodiment, when calculating the intermediate angle and performing constant false alarm rate detection based on the angle dimension fast Fourier transform, the subarray and arrangement vector can be set, and the intermediate angle of the detection target can be calculated by performing an angle dimension fast Fourier transform on the range-Doppler matrix, and inter-channel coherence accumulation can be performed; constant false alarm rate detection is performed in the velocity dimension and the angle dimension, and the detection target is precisely inspected; and super-resolution processing is performed based on the intermediate angle through a super-resolution measurement method.
[0100] In this optional embodiment, when constructing the pitch array and calculating the pitch angle and azimuth angle corresponding to the intermediate angle based on the pitch array, the sub-arrays at each pitch height can be multiplied by the steering vector of the intermediate angle and aligned to form the pitch array; the pitch array is subjected to a fast Fourier transform to obtain all pitch angles corresponding to the intermediate angle; and based on the calculated pitch angle, the azimuth angle corresponding to the intermediate angle is decoupled.
[0101] In this optional embodiment, when outputting the point cloud list based on the optimization of angle measurement accuracy and elimination of speed ambiguity, secondary interpolation processing can be performed based on the distance, speed and angle calculation results; based on the directional pattern beamforming technology, the angle measurement accuracy of the large field of view angle can be calibrated; through frequency hopping decoupling and dual-frequency wave matching, speed ambiguity can be eliminated and the point cloud list can be output.
[0102] In this optional embodiment, the point cloud list of the radar can be synchronized with the vehicle body information, and based on secondary clustering, the vehicle track can be associated and managed. When the target classification result is output, the vehicle body information can be synchronized to the radar to correct the radar measured value; through preprocessing, outliers and false points are eliminated, and point cloud classification is performed in combination with the vehicle body information; a clustering radius is set, and through a clustering, the detection targets that do not meet the clustering requirements are eliminated; based on the detection targets of the previous frame, the state of the current frame is predicted, and the point tracks are associated with the tracks; and based on the clustering algorithm, point tracks are clustered. If there is a track in the clustering result, a new track is not started; the state transition and judgment of the track are managed, and the target type is output based on the target classification result.
[0103] Figure 2 An embodiment of an automobile millimeter-wave radar signal processing system of the present invention is shown.
[0104] In this optional embodiment, the automotive millimeter wave radar signal processing system includes:
[0105] The primary signal processing module 201 is used to generate a range-Doppler data matrix through triangular pulse sampling, and convert it into a radar data cube based on range-dimensional fast Fourier transform and data compression;
[0106] The secondary signal processing module 203 is used to select radar data cubes in batches for decompression based on the distance cell distribution, and output a point cloud list through multi-dimensional filtering, phase compensation and constant false alarm rate detection;
[0107] The data processing module 205 is used to synchronize the radar point cloud list with the vehicle body information, perform preprocessing and point cloud classification, and associate and manage vehicle tracks based on secondary clustering to output target classification results.
[0108] The conventional radar signal processing is now compared and analyzed with the radar signal processing flow of the present invention.
[0109] like Figure 3 As shown in the figure, the signal processing flow of traditional radar can be summarized as follows:
[0110] (1) Range FFT: Fast Fourier Transform in the Range Dimension. Perform FFT (Fast Fourier Transform) on the chirp signal of each MIMO receiving channel to calculate the distribution of the target in each range bin (range cell), and complete the coherent accumulation in the range dimension at the same time; (2) DopplerFFT: Fast Fourier Transform in the Doppler / velocity dimension. Perform FFT between the chirp signals of each MIMO receiving channel to calculate the distribution of the target in each speed bin (speed cell), and complete the coherent accumulation of the speed dimension at the same time; (3) Win_V: Add a window (Hanning / Hamming window) to the speed dimension to reduce the impact of the side lobes in the speed dimension and reduce the probability of false alarm; (4) Win_R: Add a window (Hanning / Hamming window) to the distance dimension to reduce the impact of the side lobes in the distance dimension and further reduce the probability of false alarm; (5) Non-coherent integration: Perform non-coherent accumulation between MIMO channels, and sum the amplitudes of the data with the same distance and speed on each MIMO channel to improve the signal-to-noise ratio; (6) Doppler CFAR: After summing (non-coherent accumulation) the RD diagrams (Range-Doppler diagrams, a graphical representation method for displaying target distance and speed information) of each MIMO channel, perform CFAR (constant false alarm rate) on the synthesized RD diagram in the speed dimension; (7) Range CFAR: Perform CFAR again on the synthesized RD diagram in the distance dimension; (8) Velocity ambiguityresolution: velocity deambiguation; (9) Phase compensation: phase compensation of MIMO channel; (10) DBF: digital beamforming method to measure azimuth; (11) ElevationDOA: coherent angle measurement method or DBF method (directional pattern beamforming) to measure elevation; (12) angle matching: azimuth and elevation angle matching; (13) interpolation: interpolation to improve the accuracy of ranging, speed and angle measurements; (14) Cloudpoints list: forming a point cloud list.
[0111] The data processing flow of traditional radar is:
[0112] (1) Synch: synchronization with vehicle body information; (2) Preprocess: preprocessing - eliminating outliers / obvious false points; (3) points classification: point cloud classification (dynamic or static, high or low, etc.); (4) cluster: clustering, generally using the DBScan algorithm (density-based spatial clustering algorithm); (5) Prediction & association: point track prediction and association, generally using the extended Kalman filter; (6) Track management: track management.
[0113] Step (14) of the signal processing flow can output point cloud information to the outside, and step (6) of the signal processing can output track information to the outside.
[0114] like Figure 4 As shown, the signal processing and data processing flow of the 4D imaging radar of the present invention has two upper boxes for signal processing and the lower box for data processing. The signal processing flow of the 4D imaging radar of the present invention is:
[0115] (1) Range FFT: Fast Fourier Transform in the Range Dimension. Perform FFT (Fast Fourier Transform) on the chirp (triangular pulse) signal of each MIMO receiving channel (wireless communication channel) to calculate the distribution of the target in each range bin (range cell), and complete the coherent accumulation in the range dimension at the same time.
[0116] The FMCW waveform (Frequency Modulated Continuous Wave) received by each MIMO channel of the radar is as follows: Figure 5 As shown in Figure 1, the distance dimension information is contained in the sampling points on each chirp (triangular pulse). After sampling all the chirps in a frame, a two-dimensional matrix is obtained as follows: Figure 6 Each column of data corresponds to a chirp sampling point. Performing FFT on each column of this matrix is Range FFT, and the calculated result is the distribution of the target in each distance bin (distance cell). Figure 6 Take the first column (representing the N1 sampling points of the first chirp) as an example: perform FFT on the first column of data, and the results obtained cover the first column in order. The first row of data in the first column corresponds to the amplitude and phase of the target at a distance of 1*δR. The second row of data in the first column corresponds to the amplitude and phase of the target at a distance of 2*δR. The Nth row of data in the first column corresponds to the amplitude and phase of the target at a distance of N*δR. In this way, the distance distribution of the first chirp is obtained. δR is the value of the distance resolution, which is equal to c / (2B), c is the speed of light, and B is the effective value of the radar sweep bandwidth. In this way, traverse columns 1 to N2 to obtain the distance distribution of the chirps used.
[0117] (2) Per-chirp Compression: The FFT (distance-dimensional FFT) result of each chirp (triangular pulse) is compressed by a factor of 2.
[0118] (3) RadarDataCube: Complete the compression of the previous step for all chirps (triangular pulses), and the result is a radar data cube.
[0119] (4) Decompression: Decompress the radar cube data of several consecutive range gates, for example, decompress the radar data cube of 10 range bins from bin1 to bin10.
[0120] like Figure 7 As shown, the range-Doppler matrix of a MIMO channel / unit after decompression of 10 range bins is taken, and each row represents a range bin.
[0121] (5) DopplerFFT: Doppler / velocity fast Fourier transform. The range dimension FFT result extracted in the previous step is FFTed in the velocity dimension to calculate the distribution of the target in each velocity bin (velocity cell) and obtain the range-Doppler matrix. At the same time, the coherent accumulation in the velocity dimension is also completed.
[0122] Perform FFT on each row of the decompressed range-Doppler matrix, i.e. DopplerFFT, to calculate the distribution of the target in each speed bin (speed cell). (Compression and decompression are achieved by selecting the latest chips that support this function (such as TI, NXP radar chips).
[0123] (6) SVA_V (velocity dimension apodization filter): Add an SVA (spatial apodization filter) window to the result obtained in step (5) in the velocity dimension. The SVA window is a window function newly introduced in automotive radar in recent years, which replaces the traditional Hamming and Hanning windows. While weakening the side lobes, it hardly attenuates the main lobe amplitude (the traditional Hamming and Hanning windows have a 1-1.5dB window attenuation on the main lobe) and does not expand the main lobe width, so the resolution performance is also better than the traditional Hamming and Hanning windows (the traditional window expands the main lobe by about 1-1.5 times). SVA_V (velocity dimension apodization filter) suppresses the velocity dimension side lobes.
[0124] The implementation process takes the speed dimension as an example: Figure 6 The first column of the first row is X(1), the second column is X(2), and the kth column is X(k). The corresponding X is calculated using the following apodization filter formula. w (1), X w (2), ...X w (N2). Use X w (k) replaces X(k) and fills in the first row. This is the SVA calculation process of the first row. By traversing rows 1 to N1 in this way, we can get the SVA result of the velocity dimension. Replace the "row" in the above text with "column" and N2 with N1 to get the calculation process of the distance dimension SVA. ) Among them, the apodization filter formula is:
[0125]
[0126] In the formula, X(k) is the kth input data, X(k-1) is the k-1th input data, X(k+1) is the k+1th input data, and X w(k) is the result of processing (windowing) the input data using the apodization filter formula. X(k) is the result of FFT of the input signal x(n). α is a coefficient, as shown in the following formula:
[0127]
[0128] (7) Histogram: As shown in Figure 8, all velocity values in each range unit (bin) of the range-Doppler matrix are counted by histogram, the noise level of the range unit is estimated, and finally the noise spectrum of the entire range segment is obtained. (Statistics are made on the data amplitude distribution of the velocity unit under the range unit, and the amplitudes are arranged in descending order. The data amplitude of the k*N2th velocity unit is taken as the noise amplitude (noise spectrum) of the range unit. k is an adjustable parameter with a value range of 0 to 1, and is generally initially taken as 0.75. N2 is the number of velocity units. Repeat the traversal of all range units to obtain the noise spectrum distribution of the entire range segment.)
[0129] (8) Range dimension apodization filter SVA_R (range dimension apodization filter): Add an SVA (spatial apodization filter) window in the range dimension to the result obtained in step (6) to suppress the range dimension side lobes. The effect is similar to SVA_V.
[0130] (9) Coarse RD CFAR: Coarse RD CFAR is a method of roughly screening target points on the range-Doppler matrix. A lower threshold value is usually set (the histogram of the amplitude of each unit data on the range-Doppler matrix is statistically analyzed to obtain an amplitude value arranged in sequence. The t*N1*N2th value is taken as the threshold value, where N1 is the number of range units, N2 is the number of velocity units, and t is an adjustable parameter with a value range of 0.5 to 1 and an initial value of 0.9). Targets with values above this threshold are selected as coarse targets and enter the next process, while those with values below this threshold are discarded.
[0131] (10) Tx order Resolve: solves the transmission channel order, which is particularly suitable for the DDMA (Doppler division multiple access) system; taking the low-speed mode as an example, the distance dimension is 1024 points and the speed dimension is 256 points. 6 transmissions are transmitted simultaneously, and 2 empty sub-bands are inserted. The number of bins in each sub-band is 256 / 8=32. The phase shift scheme is shown in Table 1.
[0132] Table 1: Phase shifting scheme
[0133]
[0134]
[0135] Assuming vmax corresponds to speed bin 0, vmax corresponds to speed bin 255. The order in which the two empty sub-bands occupy the doppler frequency band can be solved, such as Figure 9-Figure 16 shown.
[0136] Table 2: Relationship between TX channel and empty subband on Doppler spectrum
[0137]
[0138] (11) Phase compensation: Phase compensation of MIMO channels, including initial phase calibration and one-degree calibration of large FOV azimuth angles. The steps of one-degree calibration are as follows:
[0139] (a) The host computer controls the radar target simulator to simulate a target with a fixed RCS (radar cross-sectional area) value at a fixed distance (for example, 10 dBsm); (b) The CAN communication interface issues a command to notify the turntable to start rotating from the maximum negative angle of the FOV (-60°), and at the same time notifies the radar to start transmitting and sampling; (c) The radar to be tested fixed on the turntable fixture starts transmitting from the maximum negative angle of the FOV (-60°). After completing the transmission and sampling at one angle, CAN will notify the host computer. After receiving the notification, the host computer will command the turntable to rotate 1° to the next angle. In this way, from -60° to -45°, the radar transmits and samples at each angle. The operation from 45° to 60° is the same. Finally, 1 degree 1 calibration data for a large FOV angle is obtained; (d) For these data to be processed, the data of each MIMO channel is separated, and the 2D-FFT (two-dimensional Fourier transform) data of the target at this angle is obtained after processing. The host computer normalizes the amplitude of this 2DFFT result (a complex number) to obtain is the phase of this MIMO channel, i = 1, 2, ... n, n is the number of MIMO channels. The corresponding MIMO channel phase compensation coefficient is obtained The phase compensation coefficient is written into the FLASH chip of the radar, and the data collected at all angles is traversed similarly to generate the phase compensation coefficient of each MIMO channel at the corresponding angle and store it in the flash. This achieves one-degree-one-calibration at a large FOV angle.
[0140] (12) Angle-FFT: Angle-dimensional FFT, which calculates the middle angle of the target and performs inter-channel coherence accumulation.
[0141] (a) Let the main sub-array arrangement vector be (S1, S2, S3, ... Sn), where Sn is the position of the array element. The array element spacing of the main sub-array is usually an integer multiple of half a wavelength; (b) Take the data of the same distance unit and velocity unit on the MIMO channel range-Doppler matrix corresponding to Sn (see Fig.17 ) as the data of array element Sn; (c) if there are any vacancies in the half-wavelength integer multiple positions of the vector (S1, S2, S3, ... Sn), insert 0, insert 0 at the end of the vector so that the number of elements of the vector satisfies the power of 2, and replace the remaining positions with the data of step (b); (d) perform FFT on the replaced vector to obtain the intermediate angles a1, a2, ... and so on of each target.
[0142] (13) DA CFAR: Doppler-angle CFAR, which performs CFAR in the velocity dimension and the angle dimension, and further detects the target based on the Coarse RD CFAR.
[0143] (14) DML: Maximum likelihood estimation method, a super-resolution angle measurement method, which performs super-resolution based on the target angle initially selected in step (12) to improve the angle measurement resolution.
[0144] (15) Elevation arrays Combine: The sub-arrays at each elevation height are multiplied by a steering vector at a certain intermediate angle to align them into a vertical elevation array.
[0145] Let raz represent the vector formed by the range-velocity FFT results of each array element in the main sub-array.
[0146] Let cg represent the vector g formed by the range-velocity FFT results of each array element in the subarray.
[0147] The whole process can be described as follows:
[0148] (a) Use the FFT result of the main sub-array raz to estimate the number of intermediate angles α, as detailed in step (4), let Θ be the set of intermediate angles. |Θ| is the size of Θ. (b) Calculate the results of the corresponding angle θi using the FFT of the main sub-array, as well as cg·wcg(θi) of each sub-array, and merge them into a vector rev. θi represents an element in the set Θ. wcg(θi) represents the steering vector corresponding to θi from the sub-array. (c) Use FFT to estimate the pitch angle of rev, similar to step (4). Let Φθ represent the set of pitch angles with the same intermediate angle θi. |Φθi| is the number of elements in the Φθi set.
[0149] (16) Elevation angle FFT: Perform FFT on the vertical elevation array obtained in step (15) to obtain all elevation angles corresponding to a certain intermediate angle. Steps (15) and (16) traverse all angles to obtain all intermediate angles and their corresponding elevation angles.
[0150] (17) Azimth-Elevation decoupling: The azimuth-elevation angle is decoupled from the elevation angle obtained in step (16) to obtain the azimuth angle corresponding to each intermediate angle. By knowing the intermediate angle α and the pitch angle Decouple the azimuth angle θ.
[0151] (18)Azimth angle Calculation: Azimth angle calculation; Azimth angle
[0152] (19) interpolation: interpolation operation, which is used to improve measurement accuracy. Perform secondary interpolation on the distance, speed, and angle results calculated in the previous step; taking distance dimension interpolation as an example, assume that the values of the three consecutive distance units (distance bins) calculated by the distance dimension FFT are x(k-1), x(k), and x(k+1) at the k-1, k, and k+1th order. Perform secondary parabola interpolation (or Lagrange interpolation) on these three values to obtain a curve x'(t). The peak value x'(k') of x'(t) is used to replace the value x(k), which represents the distance value after interpolation.
[0153] (20) Large angle DBF: For targets with large FOV angles, DBF is calibrated at one degree to improve the angle measurement accuracy. A brief description of the angle measurement calculation method of large FOV angle DBF is as follows: The guidance vector data in the interval (-60°, -45°), (45°, 60°) stored in the flash is used to perform vector dot multiplication with the received data of each MIMO channel of the radar. When a target exists at a certain angle, the data of each MIMO channel is dot-multiplied with the guidance vector corresponding to the angle to obtain an extreme value (the absolute value is approximately equal to the number of array elements n of the MIMO channel). At this time, it is determined that a target exists at this angle.
[0154] (21)step-Frequency RV decoupling: frequency hopping decoupling.
[0155] (22)AB wave matching: AB waves are used to resolve velocity ambiguity and the remainder theorem is used to expand the angle measurement range.
[0156] (23)Cloudpoints list: Output point cloud list.
[0157] (24) Return to step (4), take the next batch of radar cube data of distance bins and decompress them, and repeat steps (5) to (23) until all distance bins are traversed.
[0158] The data processing flow of the 4D imaging radar of the present invention is as follows:
[0159] (1) Synch: Synchronize with vehicle body information. Because the radar is installed on a moving vehicle, the target speed and angle values measured by the radar are relative values. The real-time vehicle speed and vehicle turning angle (yaw angle) are input to the radar through the vehicle body CAN. When the radar measures the target speed and angle information, it uses the real-time vehicle speed and yaw angle to correct the radar measured values and obtain the absolute value / relative value of the target speed and angle.
[0160] (2) Preprocess: Preprocessing - eliminating outliers / obvious false points, coordinate system conversion, parameter initialization, and filtering points in the unimportant interval or obvious false points based on the parameters of the point cloud.
[0161] (3) points classification: point cloud classification (dynamic, static, high, low, etc.), based on the vehicle's speed, yaw rate and other conditions, the point cloud is classified into dynamic and static, and high and low are classified based on distance and pitch angle.
[0162] (4) 1stcluster: One clustering, with a smaller cluster radius, removing close and weak points.
[0163] (5) Prediction & association: point track prediction and association, such as extended Kalman filtering, based on the state of the target in the previous frame. The state of the current frame is predicted, and the minimum neighbor algorithm is used to associate the point track and the track.
[0164] (6) 2nd cluster: Secondary clustering, which performs point track clustering based on the DBSCAN algorithm. If there are already tracks in the cluster, no new track will be started to prevent target splitting.
[0165] (7) Track management: Track management includes the transition between temporary track, confirmed track, and valid track status, as well as the determination and transition between dynamic and static track status.
[0166] (8) Objects classification: Output the target type and estimate the target type (pedestrian, electric vehicle, car, etc.) based on target speed, RCS, length and width and other parameters.
[0167] In summary, the signal processing and data processing solutions of traditional automotive radar and the 4D imaging automotive radar of this solution have the following advantages and disadvantages: Signal processing flow of traditional automotive radar: usually adopts TD MIMO system, conventional FMCW waveform, usually does not use super-resolution algorithm, window function uses Hamming / Hanning window, resulting in loss of resolution and signal-to-noise ratio, and usually uses incoherent accumulation between MIMO channels. Data processing flow of traditional automotive radar: usually no secondary clustering is performed, and no target classification is performed.
[0168] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.19 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store static information and dynamic information data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps in the above method embodiment are implemented.
[0169] Those skilled in the art will understand that Fig.19 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present invention, and does not constitute a limitation on the computer device to which the scheme of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0170] In addition, the present invention also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiment when executing the computer program.
[0171] In addition, the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiment are implemented.
[0172] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0173] The present invention is not limited to the structures which have been described above and shown in the drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.
Claims
1. A method for processing automotive millimeter-wave radar signals, characterized in that: include: The range-Doppler data matrix is generated by triangular pulse sampling and converted into radar data cube based on range-dimensional fast Fourier transform and data compression; Based on the range cell distribution, radar data cubes are selected in batches for decompression, and point cloud lists are output through multi-dimensional filtering, phase compensation and constant false alarm rate detection; The radar point cloud list is synchronized with the vehicle body information, and based on secondary clustering, the vehicle track is associated and managed to output the target classification results.
2. The automotive millimeter-wave radar signal processing method according to claim 1, characterized in that: The method of generating a range-Doppler data matrix by triangular pulse sampling and converting it into a radar data cube based on range-dimensional fast Fourier transform and data compression includes: Acquire the triangular pulses in the wireless communication channel, sample the sampling points of all triangular pulses in one frame, and obtain the range-Doppler data matrix; Perform a distance dimension fast Fourier transform on each column of the range-Doppler data matrix to calculate the distribution of the detected target in each range cell; The calculation results of the range-dimensional fast Fourier transform of the triangular pulse are compressed to obtain the radar data cube.
3. The automotive millimeter-wave radar signal processing method according to claim 2, characterized in that: Based on the distance cell distribution, radar data cubes are selected in batches for decompression, and through multi-dimensional filtering, phase compensation and constant false alarm rate detection, the output point cloud list includes: Select radar data cells of a number of continuous range cells in batches and decompress them; Perform a velocity-dimensional fast Fourier transform on the decompressed range-Doppler matrix, apply spatial apodization filtering to the calculation result, and generate a noise spectrum for each range cell based on histogram statistics; Set a threshold value, remove target points below the threshold value in the range-Doppler matrix, retain target points above the threshold value as coarsely selected targets, and proceed to the next step; Decode the transmission channel sequence and perform phase compensation on the wireless communication channel; Based on the angle-dimensional fast Fourier transform, the intermediate angle is calculated and constant false alarm rate detection is performed; Constructing an elevation array, and based on the elevation array, calculating the elevation angle and azimuth angle corresponding to the intermediate angle; Output point cloud list based on angle measurement accuracy optimization and velocity ambiguity elimination; The radar data cells of the next batch of several continuous distance cells are selected again, decompressed, and the distance, speed and angle calculations are repeated until all distance cells are traversed.
4. The automotive millimeter-wave radar signal processing method according to claim 3, characterized in that: The decompressed range-Doppler matrix is subjected to a velocity-dimensional fast Fourier transform, a spatial apodization filter is applied to the calculation result, and a noise spectrum of each range cell is generated based on histogram statistics, including: Perform a velocity-dimensional fast Fourier transform on each row of data in the range-Doppler matrix to calculate the distribution of the detected target on the velocity cell; The velocity dimension sidelobes of the velocity dimension calculation results are suppressed by using spatial apodization filtering; The histogram is used to count all the speed values in each distance cell, the noise level of the distance cell is estimated, and the noise spectrum of the entire distance cell is obtained; The range dimension apodization filtering is used to suppress the range dimension sidelobes of the velocity dimension calculation results.
5. The automotive millimeter-wave radar signal processing method according to claim 4, characterized in that: The phase compensation includes: initial phase calibration and one-degree calibration of the azimuth angle with a large field of view.
6. The automotive millimeter-wave radar signal processing method according to claim 3, characterized in that: The method of calculating the intermediate angle based on the angle-dimensional fast Fourier transform and performing constant false alarm rate detection includes: Set the sub-array and arrangement vector, calculate the middle angle of the detected target by performing angle-dimensional fast Fourier transform on the range-Doppler matrix, and perform inter-channel coherence accumulation; Perform constant false alarm rate detection in the speed dimension and angle dimension to perform precise inspection on the detection target; Through the super-resolution measurement method, super-resolution processing is performed based on the intermediate angle.
7. The automotive millimeter-wave radar signal processing method according to claim 3, characterized in that: The step of constructing the elevation array and calculating the elevation angle and the azimuth angle corresponding to the intermediate angle based on the elevation array includes: Multiply the sub-arrays at each elevation height by the steering vector of the middle angle and align them to form an elevation array; Perform fast Fourier transform on the pitch array to obtain all pitch angles corresponding to the intermediate angle; Based on the calculated pitch angle, the azimuth angle corresponding to the intermediate angle is decoupled.
8. The automotive millimeter-wave radar signal processing method according to claim 3, characterized in that: Based on the optimization of angle measurement accuracy and elimination of velocity ambiguity, the output point cloud list includes: Based on the distance, speed and angle calculation results, secondary interpolation processing is performed; Based on the pattern beamforming technology, the angle measurement accuracy of a large field of view is calibrated; Through frequency hopping decoupling and dual-frequency wave matching, velocity ambiguity is eliminated and a point cloud list is output.
9. The automotive millimeter-wave radar signal processing method according to claim 3, characterized in that: The method of synchronizing the radar point cloud list with the vehicle body information, associating and managing the vehicle track based on secondary clustering, and outputting the target classification results includes: Synchronize vehicle body information to radar and correct radar measured values; Through preprocessing, outliers and false points are removed, and point cloud classification is performed based on vehicle body information; Set a clustering radius, and remove detection targets that do not meet the clustering requirements through clustering; Based on the detected target of the previous frame, the state of the current frame is predicted, and the point traces are associated with the track. Based on the clustering algorithm, the point traces are clustered. If there is a track in the clustering result, a new track will not be started. Manage the state transition and judgment of the track, and output the target type based on the target classification results.
10. An automotive millimeter wave radar signal processing system, characterized in that: include: The first-level signal processing module is used to generate a range-Doppler data matrix through triangular pulse sampling, and convert it into a radar data cube based on range-dimensional fast Fourier transform and data compression; The secondary signal processing module is used to select radar data cubes in batches for decompression based on the distance cell distribution, and output the point cloud list through multi-dimensional filtering, phase compensation and constant false alarm rate detection; The data processing module is used to synchronize the radar point cloud list with the vehicle body information, perform preprocessing and point cloud classification, and associate and manage vehicle tracks based on secondary clustering to output target classification results.
Citation Information
Cited By
False alarm point elimination method based on radar array subarray combination
CN119719633A
Speed ambiguity resolution method suitable for 4D millimeter wave radar imaging
CN119828128A
A velocity deambiguation method for 4D millimeter-wave radar imaging
CN119828128B
A forward millimeter wave radar low speed cross target detection method and device
CN122836740A
