Multi-target detection method and system based on millimeter wave radar

Abstract

The invention discloses a multi-target detection method based on a millimeter-wave radar, and the method comprises the steps: collecting the echo data of multiple targets in a complex scene through the millimeter-wave radar, and carrying out the preprocessing of the echo data, and obtaining the distance-Doppler two-dimensional data to be processed; based on the obtained two-dimensional data to be processed, adopting a CFAR detection algorithm to obtain a coarse estimation CFAR mark detection result; the maximum module value square of each column of pulses on the two-dimensional data to be processed is stored, and a false target interference mark is obtained through dynamic adaptive threshold estimation; and based on the obtained CFAR mark and the false target interference mark, accurate detection of the target is realized through a comprehensive mark. According to the method, a CFAR-based comprehensive marking algorithm is adopted, target detection is realized by dynamically estimating background noise and a self-adaptive threshold value, multi-dimensional information such as distance, speed and time-frequency characteristics is fused, a real target and clutter are effectively distinguished, and the target can be accurately identified and related information of the target can be extracted in a complex scene.

Description

Technical Field

[0001] This invention relates to the fields of radar signal processing and multi-target detection technology, specifically to a multi-target detection method and system based on millimeter-wave radar. Background Technology

[0002] With the rapid development of intelligent transportation, autonomous driving, and security, the demand for high-precision real-time detection in complex environments is increasing. However, factors such as severe weather and insufficient lighting can significantly affect the target detection accuracy of traditional vision systems and LiDAR, especially in indoor and outdoor monitoring and intelligent security applications, potentially leading to safety hazards. Compared to traditional vision systems and LiDAR, millimeter-wave radar has significant advantages in weather adaptability, sensing range, and deployment cost, effectively compensating for these shortcomings. Traditional millimeter-wave radar target detection methods mainly rely on the constant false alarm rate (CFAR) algorithm, which adjusts the threshold to adapt to complex clutter environments, thereby improving the target detection probability. However, in multi-target detection scenarios and under low signal-to-noise ratio conditions, the performance of the CFAR algorithm often drops sharply, making it difficult to effectively handle complex target recognition tasks.

[0003] On the other hand, traditional CFAR algorithms have several limitations in terms of hardware design and deployment. First, the algorithm requires a large amount of computation and is highly dependent on storage and computing resources, making it difficult to deploy efficiently on hardware platforms. Second, the performance degradation of traditional CFAR algorithms in multi-target detection scenarios and low signal-to-noise ratio environments also limits the feasibility of achieving high-precision, low-latency detection on edge computing devices. Summary of the Invention

[0004] The present invention aims to at least partially solve the technical problems existing in the related art.

[0005] One objective of this invention is to provide a multi-target detection method based on millimeter-wave radar, which can accurately identify targets and extract their relevant information in complex scenarios.

[0006] Another objective of this invention is to provide a multi-target detection system based on millimeter-wave radar.

[0007] To achieve the above objectives, the present invention provides a multi-target detection method based on millimeter-wave radar, comprising the following steps:

[0008] S1. Use millimeter-wave radar to collect echo data of multiple targets in complex scenes, and preprocess the echo data to obtain two-dimensional range-Doppler data to be processed.

[0009] S2. Based on the obtained two-dimensional data to be processed, the CFAR detection algorithm is used to obtain a coarse estimate of the CFAR marker detection result;

[0010] S3. Store the square of the maximum modulus of each column of pulses in the two-dimensional data to be processed, and obtain the false target interference label through dynamic adaptive threshold estimation;

[0011] S4. Based on the obtained CFAR markers and false target interference markers, accurate target detection is achieved through integrated markers.

[0012] A further preferred embodiment of the present invention is that, in step S1, the echo data is preprocessed to obtain two-dimensional range-Doppler data to be processed, specifically:

[0013] S11. Select a frame of data and down-convert the echo data using a mixer. The characteristics of the intermediate frequency signal after mixer processing are as follows:

[0014]

[0015] Where L represents the millimeter radar transmitted signal, which consists of L LFM signals; T represents the sweep time of each LFM signal; B represents the signal bandwidth; and the duration of the entire measurement process is... , The time of the rising edge within a single LFM. Let c be time, and c be the speed of light. The distance between the target and the radar. For the target speed, For carrier frequency, This represents an exponential function with the natural logarithm e as its base. Represents a rectangular window function. The imaginary unit;

[0016] S12. Acquired via analog-to-digital converter at sampling frequency. After discretizing each linear frequency modulated wave, a two-dimensional FFT transform is performed to obtain the echo expression in the range-Doppler two-dimensional sampling plane:

[0017]

[0018] in p and p represent the distance dimension and the position corresponding to the Doppler sampling point, respectively. and These are respectively zero-padding The number of FFT points in the first and second dimensions;

[0019] The echo expression of the range-Doppler two-dimensional sampling plane reflects the target's range and velocity information, and the peak value of the sampling plane is found. and That is, the position of the target under test on the processing plane, and then the target distance is derived. and speed for:

[0020]

[0021] ;

[0022] S13. For the other frames of data, repeat steps S12-S13 to obtain the three-dimensional data to be processed, which consists of channel-distance-Doppler.

[0023] S14. Based on the obtained three-dimensional data to be processed, the information is fused by complex modulus squared and then real number superposition along the Rx dimension to obtain two-dimensional data to be processed by distance-Doppler.

[0024] Preferably, in step S2, based on the obtained two-dimensional data to be processed, the CFAR detection algorithm is used to obtain a coarse estimate of the CFAR marker detection result. The specific target detection decision formula is expressed as follows:

[0025]

[0026] Where Y is a random variable within the test unit, and S is the detection threshold;

[0027] The detection threshold S is calculated from the background noise estimate Z and the threshold factor. Multiplying them together, we get

[0028]

[0029] in, The squared value of the signal amplitude within the two-dimensional distance-Doppler plane is treated as a random variable, where i represents the number of points in the distance dimension and j represents the number of points in the Doppler frequency dimension; the background noise Z is all random variables within the two-dimensional reference window. The arithmetic mean.

[0030] Preferably, in step S3, the square of the maximum modulus of each pulse in the data to be processed is stored, and false target interference marks are obtained through dynamic adaptive threshold estimation. This mainly includes three types of mark generation: protection area mark generation near the maximum pulse value, large false target interference mark generation, and target protection area mark generation.

[0031] Preferably, the process of generating the protective region marker near the maximum pulse value is as follows: the maximum magnitude square of each column of pulses and its index are stored in an array. On the distance-Doppler plane, the region within a 2-unit length range around the index is multiplied by 1, and the remaining regions are multiplied by 0. The region multiplied by 1 is the protected region, denoted as FlagMaxProtect. A protective marker region is generated near the maximum magnitude square of each column of pulses.

[0032] Preferably, the process for generating the large false target interference marker is as follows:

[0033] Store the maximum value of each pulse column as the reference threshold;

[0034] Update the reference threshold according to the set working mode;

[0035] Using 0.98 times the reference threshold as the threshold for false targets, targets greater than this threshold are initially judged as false targets, while those less than this threshold are excluded. The specific formula for determining a large false target is as follows:

[0036]

[0037] Where T is the threshold value for false targets.

[0038] As a preferred option, the operating modes include:

[0039] (1) Pulse Doppler frame-by-frame processing: Select the maximum value of all reference thresholds extracted in the step as the updated reference threshold;

[0040] (2) Pulse processing: Use the reference threshold from step 1.

[0041] Preferably, the target protected area marker generation process is as follows:

[0042] A detection area is predefined on the range-Doppler plane. Targets detected within the range are marked as valid targets, while those outside the range are marked as invalid targets. The protected area is marked as FlagTgtProtect.

[0043] Preferably, step S4, based on the obtained CFAR markers and false target interference markers, achieves accurate target detection through a comprehensive marker synthesis, specifically as follows:

[0044] The obtained CFAR label is combined with the false target interference label to obtain the target detection composite label. The composite result is expressed as follows:

[0045]

[0046] in, For taking the AND symbol, To extract the negation sign;

[0047] In each independent connected region of the comprehensive label, the maximum value of each connected region is selected as the accurate detection result of the target detection, and the information of the detected multiple targets is output according to the corresponding index on the distance-Doppler two-dimensional plane.

[0048] Another aspect of the present invention provides a multi-target detection system based on millimeter-wave radar for implementing the above-described multi-target detection method, characterized in that it includes: a millimeter-wave radar, a central processing unit, a DSP core, BlockRAM, an FPGA, and a display terminal; wherein:

[0049] Millimeter-wave radar is used to collect echo data from multiple targets in complex scenarios.

[0050] The central processing unit (CPU) is mainly divided into an arithmetic logic unit (ALU), registers, and a control unit, and is used by microcontrollers to perform arithmetic and logical operations.

[0051] The DSP core is used to implement the digital signal processing operations required in multi-target detection methods, and can accelerate the processing of input echo digital signals in serial or parallel manner.

[0052] BlockRAM is used to store input radar echo data for pipelined processing in multi-target detection methods;

[0053] The FPGA is specifically a Xilinx Artix-7 FPGA, used to integrate the central processing unit, DSP core, and BlockRAM;

[0054] The display terminal is used to display information about the detected multiple targets, including target distance, speed, energy, and noise values.

[0055] Beneficial Effects: This invention improves upon the traditional CFAR algorithm, addressing the issue of drastic performance degradation in multi-target detection scenarios and under low signal-to-noise ratio conditions. It employs a comprehensive labeling algorithm based on CFAR, achieving target detection through dynamic estimation of background noise and adaptive thresholds. By fusing multi-dimensional information such as distance, velocity, and time-frequency characteristics, it effectively distinguishes real targets from clutter, accurately identifying targets and extracting relevant information in complex scenarios. The improved algorithm outputs detection labels in a pipelined parallel manner, better aligning with hardware processing logic and facilitating deployment on edge hardware devices. The system utilizes FPGA hardware that integrates both a DSP core and a CPU. During task deployment, batch digital signal calculations are performed using the DSP core, while logical judgments and other operations are handled by the CPU. This computational approach offers higher scalability and flexibility. Attached Figure Description

[0056] Figure 1The flowchart below shows the design of a multi-target detection method based on millimeter-wave radar, as provided in Example 1.

[0057] Figure 2 This is a schematic diagram of the 2D FFT transformation process of the LFMCW radar echo in Example 1.

[0058] Figure 3 This is a diagram illustrating the radar echo data input format in Example 1.

[0059] Figure 4 This is a schematic diagram of the radar echo data preprocessing process in Example 1.

[0060] Figure 5 This is a schematic diagram of the CFAR algorithm in Example 1.

[0061] Figure 6 The image shows the radar echo range-Doppler two-dimensional map and the corresponding CFAR detection result map when the target velocity is close to the maximum unambiguous velocity in Example 1.

[0062] Figure 7 The image shows the radar echo range-Doppler two-dimensional map and the corresponding CFAR detection result map in the scenario where the distance between the targets to be measured is close to the range resolution in Example 1.

[0063] Figure 8 The image shows the radar echo range-Doppler two-dimensional map and the corresponding CFAR detection result map in the scenario with an environmental signal-to-noise ratio of 5dB in Example 1.

[0064] Figure 9 This is a rendering of the target protection area generation module in Example 1.

[0065] Figure 10 This is a sample of radar echoes that form a false target in Example 1.

[0066] Figure 11 This is a second example of radar echoes that form a false target in Example 1.

[0067] Figure 12 This is a diagram showing the detection results of the CFAR-based integrated labeling algorithm in Example 1.

[0068] Figure 13 This is a diagram showing the detection results of the traditional CFAR algorithm.

[0069] Figure 14 This is a comparison chart of the detection accuracy of the algorithm before and after the improvement in the test scenario of 100 targets and SNR∈[0dB,20dB] in Example 1.

[0070] Figure 15This is a comparison chart of the detection accuracy of the algorithm before and after the improvement in the test scenario of SNR=20dB and the number of targets num∈[100,2000] in Example 1. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, embodiments of this invention, and should not be construed as limiting the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. In the description of this invention, it should be understood that the terminology used is for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0072] The following is combined with Figures 1-15 This invention describes a multi-target detection method and system based on millimeter-wave radar.

[0073] Example 1: As Figure 1 As shown, this embodiment provides a multi-target detection method based on millimeter-wave radar, including the following steps:

[0074] S1. Use millimeter-wave radar to collect echo data of multiple targets in complex scenes, and preprocess the echo data to obtain two-dimensional range-Doppler data to be processed.

[0075] S2. Based on the obtained two-dimensional data to be processed, the CFAR detection algorithm is used to obtain a coarse estimate of the CFAR marker detection result;

[0076] S3. Store the square of the maximum modulus of each column of pulses in the two-dimensional data to be processed, and obtain the false target interference label through dynamic adaptive threshold estimation;

[0077] S4. Based on the obtained CFAR markers and false target interference markers, accurate target detection is achieved through integrated markers.

[0078] In this example, the millimeter-wave radar in step S1 is specifically a linear frequency modulated continuous wave (LFMCW) radar, whose transmitted signal consists of L LFM signals. The sweep time of each LFM signal is denoted as T, and the signal bandwidth is B. The system transmits signals in a periodic repetitive manner, and the duration of the entire measurement process is [duration missing]. ,in This represents the rise time within a single LFM. At the receiver, the echo signal is first down-converted by a mixer, then acquired by an analog-to-digital converter (AD) at a sampling frequency of [frequency value missing]. The characteristics of the intermediate frequency signal after mixer processing are as follows:

[0079]

[0080] in, Let c be time, and c be the speed of light. The distance between the target and the radar. For the target speed, For carrier frequency, This represents an exponential function with the natural logarithm e as its base. Represents a rectangular window function. It is the imaginary unit.

[0081] With sampling frequency After discretizing each linear frequency modulated wave, then performing... Figure 2 The diagram illustrates a two-dimensional FFT transform. First, a time-dimensional FFT is performed on the frequency-modulated signal for each cycle, thus completing the range-dimensional frequency domain demodulation. Next, a frequency-dimensional FFT is performed on the FFT signal for multiple cycles, completing the Doppler-dimensional frequency domain solution, ultimately decoding the target's range and velocity information. The echo expression in the range-Doppler two-dimensional sampling plane is obtained as follows:

[0082]

[0083] in p and p represent the distance dimension and the position corresponding to the Doppler sampling point, respectively. and These are respectively zero-padding The number of FFT points in the first and second dimensions.

[0084] The echo expression of this range-Doppler two-dimensional sampling plane can reflect the target's range and velocity information, and the peak value of the sampling plane can be found. and That is, the position of the target under test on the processing plane, from which the target distance can be derived. and speed for:

[0085]

[0086]

[0087] The preprocessing process described above is for a single frame of data targeting a single target. Since radar echo data consists of multiple frames representing multiple targets, after the above preprocessing, the input format for each frame is a three-dimensional complex number: channel-range-Doppler (Rx-Range-Velocity). Figure 3 As shown, the input format of the millimeter-wave radar echo data used in this example after preprocessing is analyzed. Each element of the input three-dimensional complex data is a 64-bit binary complex number, with the first 32 bits being the imaginary part and the last 32 bits being the real part. The input data is arranged in the order of Velocity-Range-Rx (i.e., Rx changes first during calculation, followed by Range, and finally Velocity). The entire preprocessing flow supports discontinuous input and backpressure mechanisms.

[0088] To facilitate subsequent processing, enhance the strength of the target signal, and improve processing efficiency. For example... Figure 4 As shown, the three-dimensional input data needs to be processed by complex modulus calculation, and then real numbers are superimposed along the Rx dimension to fuse the information and obtain the two-dimensional distance-Doppler data to be processed.

[0089] In this example, the CFAR detection algorithm described in step S2 is specifically the two-dimensional cell average constant false alarm rate (2D-CA-CFAR) algorithm, the specific process of which is as follows: Figure 5 As shown. Its core task is to estimate background noise in the range-Doppler plane and calculate an adaptive threshold T to distinguish between the target signal and noise. Noise estimation depends on the design of the reference window, which typically contains M×N reference cells surrounding a single test cell. If the reference window contains the target signal, the noise estimation will be distorted. To avoid this problem, guard cells need to be placed around the test cell to eliminate interference from the target signal. Figure 5 As shown, the two-dimensional CFAR processing window uses a 5×5 reference cell and a 3×3 guard cell to ensure the accuracy of noise estimation.

[0090] Treating the square of the signal amplitude within the two-dimensional distance-Doppler plane as a random variable, using... Let represent the distance dimension, where i represents the number of points in the distance dimension and j represents the number of points in the Doppler frequency dimension. During detection, the random variable within the test unit is represented by Y, and the detection threshold S is determined by the background noise estimate Z and the threshold factor. Multiplying them together yields the result. Here, the background noise Z represents all random variables within the two-dimensional reference window (excluding the test unit). The arithmetic mean of the signal strength is used. Target detection is achieved by comparing the signal strength of the test unit with the detection threshold S.

[0091] The detection threshold S can be expressed as:

[0092]

[0093] The target detection decision formula based on the average false alarm rate of two-dimensional cells can usually be expressed as:

[0094]

[0095] The two-dimensional unit average constant false alarm rate (CFAR) detection algorithm in this example struggles to effectively filter out interference factors under conditions closer to real-world scenarios, such as increased ambient white noise, excessively far target distances, excessively high target speeds, and excessively close target proximity. This leads to false positives or false negatives in target detection, limiting its practical application performance. Consequently, it exhibits a sharp performance degradation in multi-target detection scenarios and under low signal-to-noise ratio conditions. The following section summarizes three scenarios with these interference factors and analyzes the algorithm's limitations.

[0096] according to Figures 6 to 8 The results of the CFAR algorithm in three scenarios show that the algorithm's target detection performance varies significantly across different scenarios. Figure 6 In scenario one, due to the high target velocity, approaching the maximum unambiguous velocity, the radar echo energy distribution in the velocity dimension is relatively dispersed. A flat region with a size close to the peak appears near the peak, causing the echo energy of the cut-off window (CUT) to be below the noise threshold, resulting in only four targets being detected, leading to missed detections. Figure 7 In scenario two, when the target spacing is too close, approaching the range resolution, the sidelobes of the two target echoes form local peaks at the intersection of the range and velocity dimensions, creating a false target. The algorithm misclassifies this as six targets. Figure 8 In scenario three, due to the influence of environmental noise, when the signal-to-noise ratio (SNR) is 5dB, a local double-peak region appears near the target, which also leads to the algorithm misclassifying it as six targets. These problems significantly limit the application of the traditional CFAR algorithm in real complex scenarios. Therefore, the FlagCfarDetect labeling result of formula (6) is used as the coarse estimation labeling result.

[0097] In this example, step S3 stores the square of the maximum magnitude of each pulse in the data to be processed, and obtains false target interference labels through dynamic adaptive threshold estimation. It mainly includes three label generation processes: protection area label generation near the maximum pulse value; large false target interference label generation; and target protection area label generation.

[0098] The generation of the protection region marker near the pulse maximum value includes:

[0099] When there are too many targets in the scene, sometimes the sidelobes of the echoes of two targets will form a local peak at the intersection of the range and velocity dimensions, creating a false target. The echo pattern is as follows: Figure 9As shown, the black circle area represents false targets generated by the interference of two targets. If the detection algorithm relies solely on a set threshold to eliminate interference, a false target exceeding the threshold will be identified as a real target. Therefore, it is necessary to eliminate false targets generated in this situation during the final detection and judgment process. The protection region generation module near the maximum pulse value is designed specifically for this scenario.

[0100] In the range-Doppler two-dimensional plane, a target always appears at the location of the maximum value of its pulse. Therefore, we store the maximum pulse value in each column of radar echo and set a protection marker within an N-grid rectangular area around it. An array stores the square of the maximum magnitude of each column of pulses and its index. In the range-Doppler plane, the area within N units of the index is multiplied by 1, and the remaining areas are multiplied by 0. The area multiplied by 1 is the protected area, denoted as FlagMaxProtect. When this protection marker overlaps with the marker area FlagCfarDetect generated by the CFAR algorithm, targets identified by CFAR within the overlapping area are considered valid real targets. In this way, local peaks caused by the sidelobes of two target echoes at the intersection of the range and velocity dimensions can be effectively eliminated, thereby improving the accuracy of target detection.

[0101] The generation of the large false target interference marker includes:

[0102] When the signal-to-noise ratio (SNR) is low in the detection scene, a local double-peak region may appear near the target, generating a false target. The echo pattern is as follows: Figure 10 As shown, the black circled area represents false targets caused by low signal-to-noise ratio. In radar signal processing, deceptive interference or false targets often exhibit abnormally strong signal amplitudes, which may lead to misjudgment by target detection algorithms. The large false target marker generation module identifies these abnormal signals by analyzing the squared magnitude data of the reference channel and generates a binarized marker matrix (FlagBigNtgtJam) to indicate which pulses and range gates may be affected by large false targets or interference.

[0103] The algorithm processing flow is as follows: 1. Store the maximum value of each pulse (range gate) as the reference threshold; 2. Update the reference threshold according to the set working mode. There are two main modes: one is pulse Doppler processing by frame, that is, selecting the maximum value of all reference thresholds extracted in step 1 as the updated reference threshold; the other is processing by pulse, that is, using the reference threshold from step 1; 3. Use 0.98 times the reference threshold as the threshold T for false targets. If it is greater than this threshold, it is initially judged as a false target; if it is less than this threshold, it is excluded.

[0104] The formula for determining a target as a major false alarm is as follows:

[0105]

[0106] The generation of the target protected area marker includes:

[0107] In detection scenarios with concentrated target distribution areas and high single-frame resolution, targets may be concentrated in a specific region of the velocity-Doppler map. A large portion of this region does not need to be scanned by the detection algorithm. Pre-defining the detection area can effectively avoid wasting computational resources and improve detection performance. In complex detection scenarios, it is also necessary to define the detection area to adapt to different needs. A target protection region generation module is used to pre-define the area to be detected, achieving the following effect: Figure 11 As shown, the white area represents the candidate protection area. Targets detected within this area will be considered valid targets and processed further. The black area outside the white area is defined as the unprotected area, where the detection algorithm will not perform target detection operations.

[0108] A detection area is pre-defined on the range-Doppler plane. Targets detected within the area are marked as valid targets, while those outside the area are considered invalid. The protected area is marked as FlagTgtProtect. This effectively avoids wasting computing power and improves detection performance. In complex detection scenarios, it is also necessary to define the detection area to adapt to different needs.

[0109] In this example, step S4 integrates the CFAR markers with the false target interference markers to achieve accurate target detection, specifically including:

[0110] The tags generated in steps S2 and S3 are combined to obtain the target detection composite tag. The composite result is shown below:

[0111] (8)

[0112] in For taking the AND symbol, To take the negation sign.

[0113] In each independent connected region of the comprehensive label, the maximum value of each connected region is selected as the accurate detection result of the target detection, and the information of the detected multiple targets is output according to the corresponding index on the distance-Doppler two-dimensional plane.

[0114] To verify the effectiveness of this invention, simulation experiments were conducted using MATLAB to validate the proposed method. The experimental results show the effectiveness of the proposed method, and the simulation results are described below:

[0115] Simulation parameters: carrier frequency speed of light ,wavelength ,bandwidth Pulse duration Antenna spacing Target angle Distance dimensional points Speed ​​dimension points .

[0116] This invention collects the echo information of the target scanned by the linear frequency modulated signal through millimeter-wave radar, and adopts a comprehensive labeling algorithm based on CFAR. It achieves target detection by dynamically estimating background noise and adaptive threshold, and integrates multi-dimensional information such as distance, velocity, and time-frequency characteristics to effectively distinguish between real targets and clutter, and outputs information on the distance, velocity, energy and noise value of the detected target. Figure 12 When the number of targets to be tested is 100, the output results displayed on the terminal of the improved algorithm of this invention are as follows: the black part is the background information that has been filtered out, the white part is the connected region of the comprehensive mark, and the green dot is the real target detected. Figure 13 The two results are the output displayed on the terminal for the traditional Cfar algorithm under the same conditions. Comparing the two results, it can be seen that the improved algorithm selects a smaller range in the white connected region, indicating that the improved algorithm selects a more accurate range of the region before marking the target, which is more conducive to the final detection and marking of the target point.

[0117] To analyze the robustness and multi-target detection capabilities of the algorithm, this example sets up two multi-target detection scenarios to test the algorithm's detection accuracy. One scenario involves setting the number of detected targets to 100 and changing the environmental SNR ∈ [0dB, 20dB]. The test results are as follows: Figure 14 As shown, compared to the traditional Cfar algorithm, the improved algorithm has a higher overall test curve and achieves a detection accuracy of 92% at SNR=11dB, indicating that the improved algorithm has a significantly improved anti-interference capability. The second scenario involves setting the environment SNR=20dB and changing the number of detected targets num∈[100,2000]. The test results are as follows. Figure 15 As shown, compared to the traditional Cfar algorithm, the improved algorithm has a higher overall test curve, and has a detection accuracy of nearly 99% when num=100, and a detection accuracy of 92% when num=1800, indicating that the improved algorithm has also greatly improved its multi-target detection capability.

[0118] Example 2: A multi-target detection system based on millimeter-wave radar, used to implement the above-mentioned multi-target detection method, includes: millimeter-wave radar, central processing unit, DSP core, BlockRAM, FPGA, and display terminal; wherein:

[0119] Millimeter-wave radar is used to collect echo data from multiple targets in complex scenarios.

[0120] The central processing unit (CPU) is mainly divided into an arithmetic logic unit (ALU), registers, and a control unit, and is used by microcontrollers to perform arithmetic and logical operations.

[0121] The DSP core is used to implement the digital signal processing operations required in multi-target detection methods, and can accelerate the processing of input echo digital signals in serial or parallel manner.

[0122] BlockRAM is used to store input radar echo data for pipelined processing in multi-target detection methods;

[0123] The FPGA is specifically a Xilinx Artix-7 FPGA, used to integrate the central processing unit, DSP core, and BlockRAM;

[0124] The display terminal is used to display information about the detected multiple targets, including target distance, speed, energy, and noise values.

[0125] In this embodiment, the millimeter-wave radar system transmits millimeter-wave signals and receives echo data. It detects the position, velocity, and other characteristics of objects by measuring changes in signal reflection, enabling high-precision detection of short-range and relatively small objects. The central processing unit (CPU) is responsible for executing program instructions, performing logical judgments and controlling system operations, handling basic arithmetic, managing data flow and control signals between modules, coordinating the operation of the entire system, and executing different algorithms and processing tasks to adapt to different detection needs. The DSP core is a computing unit specifically designed for processing digital signals, capable of rapidly executing complex calculations and processing tasks related to multi-target detection. It has parallel processing capabilities and can process multiple targets simultaneously. Complex signal processing in the data stream, optimized for specific algorithms, significantly improves computation speed and shortens target detection latency; BlockRAM memory supports double data transmission, transmitting data simultaneously on both the rising and falling edges of each clock cycle, greatly increasing data transmission rate, providing fast data storage and access capabilities, avoiding data processing bottlenecks, and improving overall system performance; FPGA integrates CPU, DSP cores, and BlockRAM, and through the parallel processing capabilities of FPGA, it can effectively execute complex signal processing algorithms, achieving parallel signal processing and shortening response time; the display terminal is used to present the real-time calculation results of multi-target detection, visualizing the processed data.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-target detection method based on millimeter-wave radar, characterized in that, Includes the following steps: S1. Use millimeter-wave radar to collect echo data of multiple targets in complex scenes, and preprocess the echo data to obtain two-dimensional range-Doppler data to be processed. S2. Based on the obtained two-dimensional data to be processed, the CFAR detection algorithm is used to obtain a coarse estimate of the CFAR marker detection result; S3. Store the square of the maximum modulus of each column of pulses in the two-dimensional data to be processed, and obtain the false target interference label through dynamic adaptive threshold estimation; S4. Based on the obtained CFAR markers and false target interference markers, accurate target detection is achieved through integrated markers.

2. The multi-target detection method based on millimeter-wave radar according to claim 1, characterized in that, In step S1, the echo data is preprocessed to obtain two-dimensional range-Doppler data to be processed, specifically as follows: S11. Select a frame of data and down-convert the echo data using a mixer. The characteristics of the intermediate frequency signal after mixer processing are as follows: ； Where L represents the millimeter-wave radar transmitted signal, which consists of L LFM signals; T represents the frequency sweep time of each LFM signal; B represents the signal bandwidth; and the duration of the entire measurement process is... , The time of the rising edge within a single LFM. Let c be time, and c be the speed of light. The distance between the target and the radar. For the target speed, For carrier frequency, This represents an exponential function with the natural logarithm e as its base. Represents a rectangular window function. The imaginary unit; S12, via analog-to-digital converter, at sampling frequency Each linear frequency modulated wave is discretized, and then a two-dimensional FFT transform is performed to obtain the echo expression in the range-Doppler two-dimensional sampling plane: ； in p and p represent the distance dimension and the position corresponding to the Doppler sampling point, respectively. and After padding with zeros The number of FFT points in the first and second dimensions; The echo expression of the range-Doppler two-dimensional sampling plane reflects the target's range and velocity information, and the peak value of the sampling plane is found. and That is, the position of the target under test on the processing plane, and then the target distance is derived. and speed for: ； ； S13. For the other frames of data, repeat steps S12-S13 to obtain the three-dimensional data to be processed, which consists of channel-distance-Doppler. S14. Based on the obtained three-dimensional data to be processed, the information is fused by complex modulus squared and then real number superposition along the Rx dimension to obtain two-dimensional data to be processed by distance-Doppler.

3. The multi-target detection method based on millimeter-wave radar according to claim 2, characterized in that, In step S2, based on the obtained two-dimensional data to be processed, the CFAR detection algorithm is used to obtain a coarse estimate of the CFAR marker detection result. The specific target detection decision formula is expressed as follows: ； Where Y is a random variable within the test unit, and S is the detection threshold; The detection threshold S is calculated from the background noise estimate Z and the threshold factor. Multiplying them together, we get ； in, The squared value of the signal amplitude within the two-dimensional distance-Doppler plane is treated as a random variable, where i represents the number of points in the distance dimension and j represents the number of points in the Doppler frequency dimension; the background noise Z is all random variables within the two-dimensional reference window. The arithmetic mean.

4. The multi-target detection method based on millimeter-wave radar according to claim 3, characterized in that, In step S3, the square of the maximum modulus of each pulse in the data to be processed is stored. The false target interference mark is obtained by dynamic adaptive threshold estimation. This mainly includes three mark generation processes: mark generation of the protected area near the maximum pulse value, mark generation of large false target interference mark, and mark generation of the target protected area.

5. The multi-target detection method based on millimeter-wave radar according to claim 4, characterized in that, The process of generating the protection region marker near the maximum pulse value is as follows: the maximum magnitude square of each column of pulses and its index are stored in an array. On the distance-Doppler plane, the region within a 2-unit length range around the index is multiplied by 1, and the remaining regions are multiplied by 0. The region multiplied by 1 is the protected region, denoted as FlagMaxProtect. A protection marker region is generated near the maximum magnitude square of each column of pulses.

6. The multi-target detection method based on millimeter-wave radar according to claim 5, characterized in that, The process of generating large false target interference markers is as follows: Store the maximum value of each pulse column as the reference threshold; Update the reference threshold according to the set working mode; Using 0.98 times the reference threshold as the threshold for false targets, targets greater than this threshold are initially judged as false targets, while those less than this threshold are excluded. The specific formula for determining a large false target is as follows: ； Where T is the threshold value for false targets.

7. The multi-target detection method based on millimeter-wave radar according to claim 6, characterized in that, The set working modes include: (1) Pulse Doppler frame-by-frame processing: Select the maximum value of all reference thresholds extracted in the step as the updated reference threshold; (2) Pulse processing: Use the reference threshold from step 1.

8. The multi-target detection method based on millimeter-wave radar according to claim 7, characterized in that, The process for generating the target protected area marker is as follows: A detection area is predefined on the range-Doppler plane. Targets detected within the range are marked as valid targets, while those outside the range are marked as invalid targets. The protected area is marked as FlagTgtProtect.

9. The multi-target detection method based on millimeter-wave radar according to claim 8, characterized in that, Step S4, based on the obtained CFAR markers and false target interference markers, achieves accurate target detection through comprehensive marker integration, specifically as follows: The obtained CFAR label is combined with the false target interference label to obtain the target detection composite label. The composite result is expressed as follows: ； in, For taking the AND symbol, To extract the negation sign; In each independent connected region of the comprehensive label, the maximum value of each connected region is selected as the accurate detection result of the target detection, and the information of the detected multiple targets is output according to the corresponding index on the distance-Doppler two-dimensional plane.

10. A multi-target detection system based on millimeter-wave radar, used to implement the multi-target detection method according to any one of claims 1 to 9, characterized in that, include: Millimeter-wave radar, central processing unit, DSP core, BlockRAM, FPGA, and display terminal; among which: Millimeter-wave radar is used to collect echo data from multiple targets in complex scenarios. The central processing unit (CPU) is mainly divided into an arithmetic logic unit (ALU), registers, and a control unit, and is used by microcontrollers to perform arithmetic and logical operations. The DSP core is used to implement the digital signal processing operations required in multi-target detection methods, and can accelerate the processing of input echo digital signals in serial or parallel manner. BlockRAM is used to store input radar echo data for pipelined processing in multi-target detection methods; The FPGA is specifically a Xilinx Artix-7 FPGA, used to integrate the central processing unit, DSP core, and BlockRAM; The display terminal is used to display information about the detected multiple targets, including target distance, speed, energy, and noise values.

Images