A millimeter wave radar-based obstacle detection method, system, and device

Abstract

The application provides a millimeter wave radar-based obstacle detection method, system and device, the method comprising: obtaining original sampling data output by each receiving channel of a millimeter wave radar after analog-digital conversion; preprocessing the original sampling data to obtain dimension-reduced range-slow-time-beam data; using an adaptive virtual phase center alignment method to perform ground clutter suppression on the dimension-reduced data to obtain a clutter-suppressed signal; performing velocity dimension spectrum analysis and constant false alarm rate detection on the clutter-suppressed signal, and combining single-pulse angle measurement to extract point cloud data of a target obstacle; using a spatial clustering algorithm to cluster the point cloud data to obtain spatial position coordinates of the target obstacle. The millimeter wave radar-based obstacle detection method can significantly improve the detection capability of low-speed and stationary obstacles in a strong clutter environment, and has the advantages of strong robustness, low computational complexity and suitability for embedded implementation.

Description

Technical Field

[0001] This application belongs to the field of radar detection and signal processing technology, and relates to an obstacle detection method based on millimeter-wave radar, and particularly to an obstacle detection method, system and device based on millimeter-wave radar. Background Technology

[0002] Millimeter waves typically refer to electromagnetic waves with wavelengths between 1 and 10 mm (corresponding to frequencies of 30-300 GHz), with the 24 GHz, 60 GHz, and 77 GHz bands being the most widely used in short-range detection radar. Their wavelengths lie between microwaves and infrared, giving them both the excellent penetration of microwaves (e.g., through smoke, fog, and dust) and the high resolution potential of the infrared band. With advancements in CMOS (Complementary Metal-Oxide-Semiconductor) and MIMO (Multiple-Input Multiple-Output) array technologies, the power consumption and size of millimeter-wave radar have been significantly reduced, making it one of the core sensors for environmental perception in platforms such as small drones and autonomous vehicles.

[0003] In terms of signal modulation schemes, FMCW (Frequency-Modulated Continuous Wave) has become the mainstream choice for millimeter-wave radar due to its unique advantages. FMCW radar transmits a continuous wave signal whose frequency varies linearly with time. An intermediate frequency (IF) signal is obtained by mixing the echo with the transmitted signal. The frequency is proportional to the target distance, and the phase change is proportional to the target's radial velocity. This scheme offers advantages such as low peak transmit power, no range blind zone, and the ability to simultaneously obtain target range, velocity, and angle information in a single scan. The large bandwidth provided by the millimeter-wave band is the physical basis for FMCW's high range resolution. The combination of these two factors enables the radar to operate stably in low-visibility environments such as smoke, rain, and fog.

[0004] However, existing technologies have already explored the direct application of FMCW millimeter-wave radar to obstacle avoidance in drones. Examples include the following:

[0005] Reference 1 (Wessendorp, Nikhil, et al. "Obstacle Avoidance onboard MAVsusing a FMCW RADAR." arXiv preprint, 2 Mar. 2021, arXiv:2103.02050.) proposes a method for obstacle avoidance directly on a UAV using a 1-transmitter, 2-receiver FMCW radar. This method employs a 1-transmitter, 2-receiver FMCW radar, tilting it upwards by 10° to passively reduce ground reflections. A Doppler-range map is generated using a two-dimensional FFT (Fast Fourier Transform), and the target angle is measured using dual-antenna phase interferometry. However, this method has insufficient suppression of ground clutter: the upward tilting installation is only a passive avoidance method, resulting in weak detection capabilities below the UAV and difficulty in distinguishing ground clutter from real targets. In outdoor or low-altitude flight scenarios, ground clutter may still enter the main lobe, causing problems such as misselecting stationary ground objects as candidate points and difficulty in distinguishing low-speed targets from ground clutter.

[0006] In the field of vehicle-mounted millimeter-wave radar, reference 2 (Huang Changba. Research on target detection technology of vehicle-mounted millimeter-wave radar [D]. Chengdu: University of Electronic Science and Technology of China, 2020. DOI: 10.27005 / d.cnki.gdzku.2020.002634) proposes a detection method based on three-dimensional FFT and CFAR (Constant False Alarm Rate), and uses a deep learning model (RD-CNN) to classify the range-Doppler spectrum to improve multi-target detection capability. However, the shortcomings of this method are: the three-dimensional FFT has high computational complexity, making it difficult to balance high resolution and the processing frame rate required for real-time obstacle avoidance on embedded platforms; the RD-CNN model only relies on the amplitude information of the range-Doppler spectrum, losing the rich features contained in the phase, and also faces problems such as strong scene dependence, the need for a large amount of labeled data, and poor model interpretability, making it difficult to generalize to different scenarios.

[0007] The research on vehicle-mounted FMCW radar disclosed in Reference 3 (Zhang Haoran. Design of Vehicle-mounted Millimeter-wave Radar Obstacle Detection System [D]. Jinan: Shandong University, 2019) adopts a relatively conservative vehicle-mounted FMCW radar processing scheme. Its process is as follows: after performing range FFT on the intermediate frequency signal, CFAR is used to detect peak values ​​to obtain the target range; before performing velocity FFT, coherent phase difference method is used to filter out stationary clutter, followed by single-pulse phase angle measurement to obtain the target angle, and finally, DBSCAN clustering algorithm (Density-Based Spatial Clustering of Applications with Noise) is used to process point cloud data to detect possible targets. However, the main drawback of this method is that CFAR is only applied after range FFT rather than on the range-Doppler spectrum, resulting in the inability to distinguish targets with different speeds within the same range cell. Stationary clutter, slow-moving targets, and noise overlap in the range dimension, making it difficult to set a unified and reasonable detection threshold for CFAR. The current mainstream processing method usually applies a two-dimensional CFAR algorithm to the range-Doppler spectrum to utilize information from both the range and velocity dimensions.

[0008] In summary, existing target detection methods based on FMCW millimeter-wave radar still have significant shortcomings in terms of ground clutter suppression, computational efficiency, phase information utilization, and multi-dimensional joint detection, making it difficult to directly meet the real-time, accuracy, and robustness requirements of small UAVs for low-altitude target detection in complex environments. Summary of the Invention

[0009] This application provides an obstacle detection method, system, and device based on millimeter-wave radar, which solves the technical problems in the existing millimeter-wave radar obstacle detection process, such as insufficient ground clutter suppression capability, difficulty in balancing computational complexity and real-time performance, poor range-velocity dimension target resolution capability, and main lobe offset caused by platform motion leading to unstable detection.

[0010] In a first aspect, this application provides an obstacle detection method based on millimeter-wave radar. The method includes: acquiring raw sampling data output from each receiving channel of the millimeter-wave radar after analog-to-digital conversion; preprocessing the raw sampling data to obtain dimensionality-reduced range-slow-time-beam data; using an adaptive virtual phase center alignment method to suppress ground clutter in the range-slow-time-beam data to obtain a clutter-suppressed signal; performing velocity-dimensional spectrum analysis and constant false alarm rate detection on the clutter-suppressed signal, and combining it with monopulse angle measurement to extract point cloud data of the target obstacle; the point cloud data includes: range, velocity, and angle information; using a spatial clustering algorithm to cluster the point cloud data to obtain the spatial coordinates of the target obstacle, and finally determining the true position of the target obstacle.

[0011] In this application, range compression and channel dimensionality reduction are achieved by performing range FFT and digital beamforming on radar echo data. An adaptive virtual phase center alignment technique is employed, combined with fuselage motion state perception and LMS (Least Mean Square algorithm) adaptive filtering, to dynamically compensate for phase shifts caused by platform motion, effectively suppressing ground clutter. A range-Doppler spectrum is constructed using Doppler FFT, and a mask is generated based on the ground clutter geometry model. With the mask's assistance, constant false alarm rate (CFAR) detection is performed to extract candidate point clouds. Finally, the DBSCAN clustering algorithm is used to cluster the point clouds, eliminating false alarms and outputting the true obstacle targets. This application significantly improves the detection capability of low-speed, stationary obstacles in strong clutter environments, and has the advantages of strong robustness, low computational complexity, and suitability for embedded implementation.

[0012] In one implementation of the first aspect, preprocessing the original sampled data to obtain dimensionality-reduced range-slow-time-beam data includes: performing range-oriented compression on the original sampled data to obtain a range-dimensional data cube; and performing dimensionality reduction processing on the range-dimensional data cube using a beam signal to obtain dimensionality-reduced range-slow-time-beam data.

[0013] In one implementation of the first aspect, an adaptive virtual phase center alignment method is used to suppress ground clutter in the range-slow-time-beam data to obtain the clutter-suppressed signal. This includes: fusing multiple sensor data through the UAV's flight control system to obtain the UAV's real-time speed; performing amplitude and phase corrections based on the UAV's real-time speed, dynamically adjusting the phase compensation amount of the two-channel data, and calculating adaptive weights for virtual phase center alignment; adjusting the adaptive weights in segments according to different range units to generate fine compensation amounts corresponding to each range unit; and using a least mean square algorithm to adaptively filter the fine compensation amounts to obtain the clutter-suppressed signal.

[0014] In one implementation of the first aspect, the multiple sensors include an inertial detection unit and a global navigation satellite system; the step of fusing multiple sensor data through the UAV's flight control system to obtain the UAV's real-time speed includes: integrating time using the accelerometer inside the inertial detection unit to calculate a first velocity component; calculating a second velocity component using the global navigation satellite system; and fusing the first velocity component and the second velocity component to obtain the UAV's real-time speed.

[0015] In one implementation of the first aspect, performing velocity-dimensional spectrum analysis and constant false alarm rate (CFAR) detection on the clutter-suppressed signal, and extracting point cloud data of the target obstacle using single-pulse angle measurement, includes: performing a velocity-dimensional Fourier transform on the clutter-suppressed signal to construct a range-Doppler spectrum; using a CFAR algorithm to perform target detection on the range-Doppler spectrum to extract candidate target points; performing single-pulse angle measurement on the candidate target points to obtain the angle information of each candidate target point, and acquiring point cloud data containing distance, velocity, and angle.

[0016] In one implementation of the first aspect, a constant false alarm rate (CFAR) algorithm is used to perform target detection on the range-Doppler spectrum, and candidate target points are extracted, including:

[0017] Based on the current speed and altitude of the UAV, the region where the clutter ridge is located is calculated using the Doppler frequency formula of the stationary ground clutter, and a mask region is generated on the range-Doppler spectrum.

[0018] Several protection units and reference units are selected on both sides of the unit to be tested. The average value of the reference units on both sides is calculated. The average value is multiplied by a coefficient to obtain the basic detection threshold.

[0019] If the unit to be detected is located within the mask area, a bias constant is superimposed on the basic detection threshold to obtain the final detection threshold; if the unit to be detected is not located within the mask area, the basic detection threshold is used as the final detection threshold.

[0020] The power value of the unit to be detected is compared with the final detection threshold, and the detection value is output based on the comparison result;

[0021] include:

[0022] When the power value of the unit to be detected is greater than the final detection threshold, the unit to be detected is determined to be a candidate target point, and the power value of the unit to be detected is used as the candidate target point; when the power value of the unit to be detected is less than or equal to the final detection threshold, 0 is output, and the unit is determined to be clutter background.

[0023] In one implementation of the first aspect, a spatial clustering algorithm is used to cluster the point cloud data to obtain the spatial location coordinates of the target obstacle. This includes: using a density-based spatial clustering algorithm to cluster the point cloud, merging point clouds that are spatially and velocally adjacent into a cluster, and calculating the geometric center of each cluster, with the center representing a real physical obstacle.

[0024] Secondly, this application provides an obstacle detection system based on millimeter-wave radar, comprising:

[0025] The data acquisition module is used to acquire the raw sampled data output from each receiving channel of the millimeter-wave radar after analog-to-digital conversion;

[0026] The data preprocessing module is used to preprocess the original sampled data to obtain the dimension-reduced distance-slow-time-beam data;

[0027] The clutter suppression module is used to perform ground clutter suppression on the range-slow time-beam data using an adaptive virtual phase center alignment method to obtain the clutter-suppressed signal.

[0028] The global spectrum analysis module is used to perform velocity-dimensional spectrum analysis and constant false alarm rate detection on the clutter-suppressed signal, and to extract point cloud data of the target obstacle by combining it with single-pulse angle measurement; the point cloud data includes: distance, velocity and angle information;

[0029] The point cloud post-processing module is used to cluster the point cloud data using a spatial clustering algorithm to obtain the spatial coordinates of the target obstacle and ultimately determine the true location of the target obstacle.

[0030] Thirdly, this application provides an electronic device. The electronic device includes: a memory for storing a computer program; and a processor for executing the computer program stored in the memory to cause the electronic device to perform the obstacle detection method based on millimeter-wave radar as described in any one of the first aspects.

[0031] As described above, the obstacle detection method, system, and device based on millimeter-wave radar described in this application have the following beneficial effects:

[0032] (1) The obstacle detection method based on millimeter-wave radar provided in this application adopts adaptive virtual phase center alignment technology. By estimating the actual motion speed of the UAV platform, the phase compensation amount is dynamically adjusted to effectively suppress ground clutter interference caused by attitude changes or wind disturbances. Compared with the traditional DPCA method, the scheme of this application is not sensitive to speed fluctuations, and the clutter cancellation is more robust, significantly improving the signal-to-clutter ratio of real obstacles in low-altitude flight scenarios. At the same time, based on AVPCA (Adaptive Cross-Polarization Virtual Array), this application further introduces an adaptive filter based on the least mean square (LMS) criterion to perform closed-loop fine-tuning of residual clutter caused by insufficient sensor accuracy. By iteratively optimizing the filter weights, the cancellation residual energy is minimized, thereby improving the accuracy and stability of clutter suppression, thus adapting to different flight platforms and environmental conditions.

[0033] (2) In this application, a clutter ridge model can be constructed based on the geometric relationship between the airborne radar and ground clutter, and a mask region can be generated on the range-Doppler spectrum. CFAR detection with an offset threshold is used within the mask region, while standard CA-CFAR (Cell-Averaging Constant False Alarm Rate) is maintained for the background region. This partitioning strategy effectively avoids false alarms caused by clutter remnants, while maintaining the ability to detect slow-moving, low signal-to-noise ratio targets.

[0034] (3) In this application, spatial domain dimensionality reduction is performed by digital beamforming (DBF), which avoids the high computational overhead of three-dimensional FFT; the nonlinear function of clutter ridge is approximated by piecewise linear approximation, which reduces the real-time computation burden. The overall algorithm is mainly based on weighting, local operation and iterative update, which has good parallelism and is easy to implement high frame rate processing on embedded DSP or FPGA platforms, thus meeting the real-time requirements of UAV obstacle avoidance.

[0035] (4) This application also uses the DBSCAN density clustering algorithm to cluster the candidate point cloud, merging spatially and velocity-proximity points into a single physical target and eliminating isolated false alarm points. It can effectively solve the problem of point cloud dispersion caused by multipath reflection, sidelobe interference, etc., and output the target position more accurately and stably.

[0036] (5) The system provided in this application adopts a three-transmitter, four-receiver (3T4R) single-chip millimeter-wave radar, combined with digital beamforming technology, to achieve channel data dimensionality reduction at the receiving end. This design significantly reduces the number of RF front-ends and data interfaces, lowers system power consumption and size, and is suitable for small UAV platforms. At the same time, the system integrates a fuselage state perception module (speed, altitude, etc.) and has range segmentation fine compensation capability, which can dynamically adjust AVPCA weights and mask areas according to current flight parameters. Therefore, the system of this application can still maintain a stable detection probability in non-uniform clutter environments such as near-ground and urban canyons, effectively suppressing false alarms caused by stationary objects on the ground and low-speed clutter, and has good environmental adaptability and dynamic response capability. Attached Figure Description

[0037] Figure 1 The diagram shows an application scenario of the hardware structure corresponding to the obstacle detection method based on millimeter-wave radar described in this application.

[0038] Figure 2 The diagram shown is a schematic representation of the obstacle detection process of the millimeter-wave radar-based obstacle detection method described in this application embodiment.

[0039] Figure 3 The diagram shown is a schematic representation of the overall process of the obstacle detection method based on millimeter-wave radar as described in the embodiments of this application.

[0040] Figure 4 The diagram shown is a schematic flowchart of step S2 of the obstacle detection method based on millimeter-wave radar described in the embodiments of this application.

[0041] Figure 5 The diagram shown is a flowchart of step S3 of the obstacle detection method based on millimeter-wave radar described in this application embodiment.

[0042] Figure 6 The diagram shows an adaptive filter for the LMS in the obstacle detection method based on millimeter-wave radar described in this application embodiment.

[0043] Figure 7 The diagram shown is a flowchart of step S4 of the obstacle detection method based on millimeter-wave radar described in this application embodiment.

[0044] Figure 8 The diagram shown is a flowchart of step S42 of the obstacle detection method based on millimeter-wave radar described in this application embodiment.

[0045] Figure 9 The diagram shows the geometric relationship of stationary ground clutter in the obstacle detection method based on millimeter-wave radar described in this application embodiment.

[0046] Figure 10 The diagram shown is a schematic representation of the CA-CFAR algorithm in the obstacle detection method based on millimeter-wave radar described in this application embodiment.

[0047] Figure 11 The diagram shown is a schematic diagram of the single-pulse angle measurement principle in the obstacle detection method based on millimeter-wave radar described in the embodiments of this application.

[0048] Figure 12 The diagram shown is a schematic representation of the structural composition of the obstacle detection system based on millimeter-wave radar as described in an embodiment of this application.

[0049] Figure 13 The diagram shown is a structural schematic of the electronic device described in an embodiment of this application.

[0050] Component designation explanation

[0051] 1 obstacle detection device based on millimeter-wave radar 11 millimeter-wave radar 12 Analog-to-digital converter 13 fuselage status sensing sensor 131 Inertial detection unit 132 Global Positioning System 133 accelerometer 134 gyroscope 14 Embedded processor 21 Data acquisition module 22 Data preprocessing module 23 Clutter suppression module 24 Global Spectrum Analysis Module 25 Point cloud post-processing module 30 electronic devices 31 processor 32 memory S1~S5 step Detailed Implementation

[0052] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0053] It should be noted that, in the embodiments of this application, the words "specifically" or "for example" indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "specifically" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "specifically" or "for example" is intended to present the relevant concepts in a specific manner.

[0054] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0055] The following embodiments of this application provide an obstacle detection method, system, and device based on millimeter-wave radar, including but not limited to hardware application scenarios such as millimeter-wave radar, analog-to-digital converter, and fuselage state perception sensor. The following description will take an obstacle detection device based on millimeter-wave radar as an example.

[0056] Please see Figure 1 and Figure 2 The figures show a schematic diagram of the hardware structure application scenario corresponding to the obstacle detection method based on millimeter-wave radar described in this application and a schematic diagram of the obstacle detection process of the obstacle detection method based on millimeter-wave radar described in the embodiments of this application, respectively.

[0057] like Figure 1 and Figure 2 As shown, this embodiment provides a hardware application scenario for an obstacle detection device based on millimeter-wave radar. The specific hardware structure mainly includes: millimeter-wave radar 11, ADC (Analog-to-Digital Converter) 12, fuselage state perception sensor 13, and embedded processor 14.

[0058] In one embodiment of this application, the millimeter-wave radar 11 is preferably a three-transmitter, four-receiver single-chip millimeter-wave radar. The three-transmitter, four-receiver single-chip millimeter-wave radar is used to transmit and receive millimeter-wave FMCW signals to acquire raw echo data such as the target's distance, velocity, and angle.

[0059] The analog-to-digital converter 12 is used to convert the analog intermediate frequency signal received by the radar into a digital signal for subsequent digital signal processing.

[0060] The fuselage state perception sensor 13 includes an inertial measurement unit (IMU) 131, an accelerometer 133, a gyroscope 134, and an optional global positioning system (GPS) 132, etc., for real-time acquisition of motion parameters such as the speed, attitude angle, and acceleration of the UAV, and transmitting these parameters to the embedded processor.

[0061] The embedded processor 14 preferably uses an FPGA (Field-Programmable Gate Array) or DSP (Digital Signal Processor) chip to receive digital intermediate frequency signals and perform corresponding processing through various signal processing algorithms, such as distance Fourier transform, digital beamforming, AVPCA, adaptive filtering, CA-CFAR detection, monopulse angle measurement, and DBSCAN clustering.

[0062] Specifically, the three-transmitter, four-receiver single-chip millimeter-wave radar employs a frequency-modulated continuous wave (FMCW) system, transmitting linear frequency-modulated continuous wave signals into the detection area and simultaneously receiving target and ground echoes through four receiving channels. The four-channel analog intermediate frequency (IF) signals output by the radar are fed into an ADC, which converts the analog IF signals into digital IF signals for subsequent processing.

[0063] The embedded processor first uses multiple transmit and receive antennas with multiple inputs and multiple outputs to synthesize 12 virtual spatial channels. Then, the embedded processor performs a range-dimensional fast Fourier transform on the intermediate frequency signal of each channel to achieve range compression. Subsequently, DBF (Digital Beamforming) is performed on the four-channel data to reduce the dimensionality of the original channel data to several directional beams, thereby reducing the computational complexity of subsequent operations.

[0064] Then, adaptive virtual phase center alignment (AVPCA) clutter suppression is performed. That is, the actual moving speed and displacement deviation of the platform are estimated by the embedded processor, the phase compensation weight between the two channels is dynamically calculated, and fine compensation is performed according to the distance segment; further, the LMS adaptive filter is used to perform closed-loop fine adjustment of the residual clutter, and the distance dimension data after clutter cancellation is output.

[0065] Next, after clutter suppression is completed, full-domain spectrum analysis and auxiliary detection are performed. Specifically, first, velocity-dimensional FFT is performed on the data of each beam to construct the range-Doppler spectrum; then, based on the current altitude and velocity provided by the fuselage status sensor, the theoretical distribution area of ​​stationary ground clutter on the range-Doppler spectrum is calculated to generate a mask area; then, CA-CFAR detection is performed, and a bias constant is forcibly superimposed in the mask area to raise the detection threshold, while standard CA-CFAR is used in the non-masked area to extract candidate target points; finally, the azimuth angle of each target is calculated using the single-pulse angle measurement method.

[0066] Finally, point cloud post-processing is performed, namely: the candidate target point cloud is clustered using a density-based spatial clustering algorithm to remove isolated false alarm points, and the final obstacle position, distance, speed and angle information is output for the UAV's flight control module to make obstacle avoidance decisions.

[0067] It should be noted that those skilled in the art should understand that the specific models, connection methods, scheduling cycles of the embedded processors, and the number of filtered data, etc., of the aforementioned hardware can all be adjusted according to the actual application scenario without departing from the scope of protection of this application.

[0068] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

[0069] In summary, all the aforementioned hardware components are integrated within the UAV's onboard electronics bay. The antenna of the three-transmitter, four-receiver millimeter-wave radar is slightly tilted downwards towards the flight path to detect both obstacles ahead and the ground below. The embedded processor connects to the ADC via a high-speed parallel or serial interface, while the processor connects to the fuselage status sensing sensors via an I2C (Inter-Integrated Circuit) or SPI (Serial Peripheral Interface) bus. The entire system processes data in real-time during flight, with a frame rate of at least 20 Hz, meeting the real-time requirements for low-altitude obstacle avoidance.

[0070] The obstacle detection method based on millimeter-wave radar in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0071] When a drone flies at a low altitude of less than 20 meters, the sidelobes of the radar beam will receive a large amount of ground clutter signals. At the same time, the attitude change caused by the drone's acceleration or braking will also change the scanning direction of the radar main lobe. Both of these factors will cause ground clutter to seriously interfere with obstacle identification.

[0072] Therefore, this application designs a set of millimeter-wave radar based on a three-transmitter-four-receiver (3T4R) single-chip to realize a joint processing method of physical cancellation and algorithm compensation, thereby achieving the purpose of effectively suppressing ground clutter.

[0073] Please see Figure 3 The diagram shows the overall flow chart of the obstacle detection method based on millimeter-wave radar described in the embodiments of this application.

[0074] like Figure 2 and Figure 3 As shown, this embodiment provides an obstacle detection method based on millimeter-wave radar, the method including the following steps:

[0075] S1: Acquire the raw sampling data output from each receiving channel of the millimeter-wave radar after analog-to-digital conversion.

[0076] In one embodiment, a millimeter-wave radar transmits a frequency-modulated continuous wave signal and receives the echo signal. After analog-to-digital conversion, a multi-channel digital intermediate frequency signal is obtained, and the digital intermediate frequency signal is output as the original sampling data.

[0077] Specifically, millimeter-wave radar (FMCW system) transmits frequency-modulated continuous wave signals. Upon encountering a target, the reflected echo is mixed with the local transmitted signal to obtain a difference frequency signal, i.e., an intermediate frequency signal. The frequency of this intermediate frequency signal is proportional to the target distance. Then, an ADC is used to sample this analog intermediate frequency signal to obtain a digital intermediate frequency signal.

[0078] S2, preprocess the original sampled data to obtain the dimension-reduced distance-slow time-beam data.

[0079] The raw ADC data acquired by the radar front end is converted into range-dimensional data with clear physical meaning, and the channel dimension is compressed using digital beamforming technology to reduce the computational complexity of subsequent signal processing stages. This stage consists of two steps: range FFT compression and digital beamforming dimensionality reduction (DBF).

[0080] Please see Figure 4 The diagram shows a flowchart of step S2 of the obstacle detection method based on millimeter-wave radar described in this embodiment. Figure 4 As shown, the specific steps of S2 include:

[0081] S21, perform distance compression on the original sampled data to obtain a distance-dimensional data cube.

[0082] In one embodiment, the acquired signal is an ADC echo signal transmitted by three transmitting antennas in a time-division manner and acquired simultaneously by four receiving antennas. The signal is then transformed by a Fourier transform on the distance dimension using the TDM-MIMO method to obtain a data cube in the distance dimension, which is a complex matrix with a size of 12*N*M.

[0083] Specifically, a three-transmit, four-receive (3T4R) FMCW millimeter-wave radar antenna system is adopted. Through TDM-MIMO (Time Division Multiplexing Multiple Input Multiple Output), the combination of three transmit antennas and four receive antennas can be expanded into 12 virtual receive channels. Each virtual channel forms a two-dimensional data structure after data acquisition, where rows correspond to fast time sampling (corresponding to the range dimension) and columns correspond to slow time series (i.e., different linear frequency modulated signals chirp, corresponding to the velocity dimension).

[0084] Furthermore, to suppress spectral leakage, the fast-time sampled data of each chirp is first windowed (e.g., using a Hanning or Hamming window), and then a range FFT is performed on each chirp. In frequency-modulated continuous wave (FMCW) radar, the beat frequency obtained after mixing the echo signal with the transmitted signal is proportional to the target range; therefore, each frequency in the frequency domain after the FFT corresponds to a specific range cell. This results in complex spectral data of length N (the original number of sampling points) for each virtual receiving channel, representing the echo information of N range cells, where each complex number simultaneously contains the amplitude and phase of the echo from that range cell. Finally, after the range FFT, a three-dimensional data cube of 12 (virtual array) * N (number of range cells) * M (number of chirps) is obtained.

[0085] S22, the distance-dimensional data cube is reduced in dimension by beam signal to obtain the reduced distance-slow time-beam data.

[0086] To reduce the computational burden of subsequent space-time processing while preserving the target's directional resolution, this step employs digital beamforming technology to compress the receiving channel dimension to the number of preset beam directions.

[0087] In one embodiment, based on the aforementioned three-dimensional vector matrix, beam signals in four directions (number of beams * number of range cells * number of chirps) are output by weighting the guide vector, thereby reducing the complexity of subsequent processing.

[0088] Specifically, based on the radar's monitoring range and the target, four fixed-point wide beams are selected. Since the primary objective here is obstacle avoidance, a smaller field of view can be chosen. ): -30°, -10°, 10°, 30° (with the radar normal direction as 0°). For each desired beam pointing angle Calculate the corresponding steering vector based on the geometry of the antenna array. By weighted summing of the vectors from the 12 virtual channels, the array output is made equivalent to a beam at a certain angle. The steering vector here is directly related to the weights, and the specific formula is as follows:

[0089]

[0090] in, Represents the conjugate form of the guiding vector matrix; A complex vector matrix representing different virtual channels.

[0091] Through the two steps described above, the raw ADC data is converted into a three-dimensional complex matrix with clearly defined distance physical meaning and channel dimension compression. This data format retains the slow-time dimension information necessary for subsequent Doppler processing and clutter suppression while significantly reducing the data size, laying a favorable foundation for the next step of real-time adaptive clutter suppression.

[0092] It should be noted that the above content describes the specific implementation process of this step. Those skilled in the art can adjust the above embodiments according to actual radar parameters (such as the number of channels, the number of beams, the beam pointing angle, etc.) without departing from the core idea of ​​the present invention.

[0093] S3, an adaptive virtual phase center alignment method is used to suppress ground clutter in the range-slow time-beam data to obtain the clutter-suppressed signal.

[0094] This process employs an adaptive virtual phase center alignment method. By estimating the platform's actual movement speed, this method dynamically adjusts the phase compensation of the two channels, making the clutter cancellation effect less sensitive to speed fluctuations.

[0095] In other words, based on the foregoing, adaptive virtual phase center alignment is performed within the four beams, specifically by processing adjacent pulses within the same beam to update the real-time weights. This process is an equal-length transformation, which does not change the Doppler sampling rate or the number of pulses accumulated in the input signal, thus ensuring the highest speed resolution for subsequent 2D-FFT processing.

[0096] Please see Figure 5The diagram shows a flowchart of step S3 of the obstacle detection method based on millimeter-wave radar described in this embodiment. Figure 5 As shown, the specific steps of S3 include:

[0097] The S31 uses the drone's flight control system to integrate data from multiple sensors to obtain the drone's real-time speed.

[0098] In one embodiment, the multiple sensors include an inertial detection unit (IMU) and a global navigation satellite system (GNSS). A first velocity component is calculated by integrating time using the accelerometer inside the IMU; a second velocity component is calculated using the GNSS; and the first and second velocity components are fused to obtain the real-time velocity of the UAV.

[0099] Specifically, the distance between the phase centers of the antenna ideally corresponds to the product of the flight speed and the pulse repetition interval. However, precise alignment is often not achieved due to variations in flight speed. Therefore, in this step, the UAV's flight control system first fuses data from various sensors to obtain the real-time speed, such as the inertial navigation system (IMU / GNSS). The IMU (Inertial Detection Unit) can perform time integration using its internal accelerometer, and the GNSS (Global Navigation Satellite System) can also calculate the speed. In other words, the UAV's flight control system fuses data from multiple sensors to obtain the UAV's real-time speed. The inertial detection unit obtains the speed through time integration using its internal accelerometer, while the global navigation satellite system obtains the speed by calculating the Doppler frequency shift or position difference.

[0100] In summary, by sensing the fuselage status, accurate and robust platform motion velocity information is provided for the entire adaptive virtual phase center alignment process, which is a prerequisite for achieving effective suppression of subsequent clutter.

[0101] S32, based on the real-time speed of the UAV, perform amplitude and phase correction, dynamically adjust the phase compensation amount of the two-channel data, and calculate the adaptive weight for virtual phase center alignment.

[0102] In one embodiment, adaptive weight calculation is performed based on the deviation to dynamically adjust the phase compensation amount of each channel data and generate preliminary phase compensation weights.

[0103] Specifically, a complex weighting factor is constructed to correct the amplitude and phase of the signal in the secondary channel.

[0104] The formula for correcting the adaptive weights is:

[0105]

[0106]

[0107]

[0108]

[0109] in, Represents the complex weighting factor; This represents the amplitude scaling factor, which corrects the gain imbalance between the receiving and transmitting channels. It is obtained through testing with the platform stationary. Indicates compensation for phase difference; This represents the complex value of the i-th distance gate in the n-th chirp of the coherent source used as a reference. This represents the secondary channel of the compensation signal, whose phase center theoretically will coincide with the signal after one pulse repetition. The phase centers coincide; Indicates wavelength; Indicates the downward angle; Indicates the pulse repetition period; This represents the difference between the ideal displacement and the actual displacement of the airborne radar during its movement. This indicates the speed at which the radar moves.

[0110] S33, the adaptive weights are adjusted in segments according to the different distance units to generate fine compensation amounts corresponding to each distance unit.

[0111] In one embodiment, the preliminary phase compensation weight is applied to the distance segmentation fine compensation, the echo data of different distance units are phase corrected respectively, and the multi-channel signal after segmentation compensation is output.

[0112] Specifically, considering the depression angle of different distance units It is a function of distance ( Therefore, the compensation weights must be dynamically adjusted according to the distance cells. Considering the balance between calculation accuracy and real-time performance, interval segmentation adjustment of parameters is used for compensation. Different complex weights ω(R) are applied to near and far cells respectively to solve the problem of missed protection that cannot be balanced between near and far distances.

[0113] Accordingly, the formula for the segmented adjustment includes:

[0114]

[0115] The formula for segmentation based on the relationship between H and R is:

[0116]

[0117] When the target distance is far relative to the UAV altitude (R>5H), cosθ=1 is directly adopted for approximate calculation; when 2H< R ≤5H, the corresponding depression angle is calculated by using the midpoint of the interval length of every 16 range bins; when R≤2H, the corresponding depression angle is calculated by using the midpoint of the interval length of every 4 range bins.

[0118] S34, a least mean square algorithm is adopted to perform adaptive filtering on the fine compensation amount, and obtain a signal after clutter suppression.

[0119] Please refer to Figure 6 , which is a schematic diagram of an LMS adaptive filter of the millimeter-wave radar-based obstacle detection method according to an embodiment of the present application.

[0120] In an embodiment, the multi-channel signals after segmented compensation are used as input and sent to an adaptive filter based on least mean square (LMS), filter coefficients are dynamically adjusted based on the criterion of minimizing the mean square error between the filter output and a reference signal, and the final signal after residual clutter suppression is output.

[0121] Specifically, for residual clutter caused by insufficient sensor accuracy, a filter is introduced for adaptive fine-tuning. The LMS method adjusts the weight to an optimal value such that the output of the filter is as close to the reference value as possible, and this degree of proximity is measured by the mean square error.

[0122] The weight update formula in the adaptive filtering is:

[0123]

[0124]

[0125] wherein, represents the iteration step size; represents the cancellation residual; represents the iteration weight parameter.

[0126] Figure 6 in the LMS adaptive filter structure of , is used as a reference input, which is multiplied by through the filter, the output is differenced with to obtain a drive signal, which is used to drive to update.

[0127] S4, performing velocity dimension spectrum analysis and constant false alarm rate detection on the signal after clutter suppression, and extracting point cloud data of a target obstacle in combination with monopulse angle measurement; the point cloud data comprises: distance, velocity and angle information.

[0128] Please see Figure 7 The diagram shows a flowchart of step S4 of the obstacle detection method based on millimeter-wave radar described in this embodiment. Figure 7 As shown, the specific steps of S4 include:

[0129] S41, Based on the clutter-suppressed signal, a velocity-dimensional Fourier transform is performed to construct a range-Doppler spectrum.

[0130] In one embodiment, firstly, a Fast Fourier Transform (FFT) is performed on the slow time series of each range cell within each beam after clutter suppression to obtain a complex spectrum in the Doppler domain. Then, the Doppler complex spectra of all range cells under each beam are organized according to the range dimension and the Doppler dimension to form the range-Doppler spectrum of that beam. In this spectrum, each element is represented by the square of the modulus of the complex spectrum (i.e., the power value), or the original complex value is retained for subsequent phase processing. Next, the generated range-Doppler spectrum data is stored in beam order, using a Doppler index-first or range index-first layout to improve data access efficiency. Finally, the constructed range-Doppler spectrum data for each beam is output to the constant false alarm rate (CFAR) detection module as input data for candidate target extraction.

[0131] S42, The constant false alarm rate algorithm is used to perform target detection on the range-Doppler spectrum and extract candidate target points.

[0132] Please see Figure 8 , Figure 9 and Figure 10 The diagrams shown are, respectively, a flowchart of step S42 of the obstacle detection method based on millimeter-wave radar described in this application embodiment, a schematic diagram of the geometric relationship of stationary ground clutter in the obstacle detection method based on millimeter-wave radar described in this application embodiment, and a schematic diagram of the CA-CFAR algorithm in the obstacle detection method based on millimeter-wave radar described in this application embodiment. Figure 8 As shown, the specific steps of S42 include:

[0133] S421, based on the current speed and altitude of the UAV, calculate the region where the clutter ridge is located using the Doppler frequency formula of the stationary ground clutter, and generate a mask area on the distance-Doppler map;

[0134] S422, Select several protection units and reference units on both sides of the unit to be detected, calculate the average value of the reference units on both sides, and multiply the average value by a coefficient to obtain the basic detection threshold.

[0135] S423, if the unit to be detected is located within the mask area, a bias constant is superimposed on the basic detection threshold to obtain the final detection threshold; if the unit to be detected is not located within the mask area, the basic detection threshold is used as the final detection threshold.

[0136] S424, compare the power value of the unit to be detected with the final detection threshold, and output a detection value based on the comparison result; including: when the power value of the unit to be detected is greater than the final detection threshold, the unit to be detected is determined to be a candidate target point, and the power value of the unit to be detected is used as a candidate target point; when the power value of the unit to be detected is less than or equal to the final detection threshold, output 0, and determine that the unit is clutter background.

[0137] In one embodiment, two-dimensional CA-CFAR detection is performed on the power spectrum of each beam direction to achieve spatially selective target detection.

[0138] Specifically, based on the current speed and altitude of the UAV, the region where clutter ridges should theoretically exist is calculated and marked as a mask region on the Doppler-range map.

[0139] For airborne radar, the formula for the Doppler frequency corresponding to clutter points on stationary ground is:

[0140]

[0141] in, This indicates the angle between the direction of the velocity and the direction of the target. This indicates the downward angle, which is the angle between the radar's line of sight to the target point and the horizontal plane; This represents the angle between the projection of the UAV's velocity direction and the direction pointing towards the target onto the horizontal plane. Specific angles are as follows: Figure 9 As shown, considering the geometric relationship of the top-down view, the above formula is improved as follows:

[0142]

[0143] in, Indicates the altitude of the drone. This indicates the slant range between the radar and the target.

[0144] This formula is a nonlinear function. To reduce the complexity of real-time computation, the entire distance dimension is divided into several sub-intervals. Within each interval, a local linear approximation of the function is made, followed by a first-order Taylor expansion, as shown in the formula:

[0145]

[0146] Furthermore, CA-CAFR (Constant False Alarm Rate Algorithm with Average Cell Count) is used for target detection across the entire domain. For example... Figure 10 As shown, n reference cells are taken on each side of the detected cell D, and their average values ​​X and Y are calculated to estimate the background power of cell D. The average value of the reference cells on both sides of the detected cell is multiplied by a coefficient to obtain the detection threshold S of the detected cell. This threshold S is compared with the detected cell D. If the detected cell D is greater than the detection threshold S, the value of the detected cell D is output; otherwise, the cell is considered clutter background, and the output result is 0. In addition, sometimes the target may be much larger than the resolution accuracy and occupy multiple cells, so k guard cells are left on each side of D. CA-CFAR is still used for the mask area, but an offset constant is forcibly superimposed to appropriately increase the detection threshold of the low clutter region. This ensures that even if there is still a slight residual after cancellation, it will not cause a large-area false alarm, while still being able to detect slow targets whose signal-to-clutter ratio has been improved.

[0147] In other words, for the unit D to be detected, a protection unit and a reference unit are selected on both sides of it respectively. The average value X of the left reference unit and the average value Y of the right reference unit are calculated. The average value is multiplied by a coefficient to obtain the basic detection threshold.

[0148] If the unit D to be detected is located within the masked area, a bias constant is superimposed on the basic detection threshold to obtain the final detection threshold S. If it is located outside the masked area, the basic detection threshold is directly used as the final detection threshold S. The power value of the unit D to be detected is compared with the final detection threshold S, and the detection value is output based on the comparison result. Specifically, if the power value of the unit D to be detected is greater than the final detection threshold S, the power value of the unit D to be detected is output as a candidate target point; if the power value of the unit D to be detected is less than or equal to the final detection threshold S, the output is 0, and the unit is determined to be clutter background.

[0149] Furthermore, repeat the above steps for all cells in the distance-Doppler graph from the preceding steps, and finally, obtain the candidate target point cloud.

[0150] S43, perform single-pulse angle measurement on the candidate target points to obtain the angle information of each candidate target point, and acquire point cloud data containing distance, velocity and angle.

[0151] Please see Figure 11 The diagram shows the principle of single-pulse angle measurement in the obstacle detection method based on millimeter-wave radar described in the embodiments of this application.

[0152] In one embodiment, a phase-based angle measurement method is used, which calculates the angle by utilizing the phase difference between the echo signals received by multiple antennas. For example... Figure 11As shown, the distance difference caused by different angles will be reflected in the phase difference of the echo. Therefore, the angle of the target is calculated based on this characteristic. The formula used is:

[0153]

[0154] in, Indicates the radar wavelength; Indicates the azimuth of the target.

[0155] It should be noted that in practical applications, the inaccuracy of signal measurement needs to be considered. Therefore, the sum-difference ratio of the two echo signals is generally used for calculation, i.e., by... Use ratios to estimate target azimuth:

[0156]

[0157] in, It represents the imaginary unit in complex number operations.

[0158] S5. A spatial clustering algorithm is used to cluster the point cloud data to obtain the spatial coordinates of the target obstacle.

[0159] In one embodiment, a density-based spatial clustering algorithm is used to cluster point clouds, merging point clouds that are spatially and velocally adjacent into a cluster, and calculating the geometric center of each cluster to represent a real physical obstacle.

[0160] Specifically, the point cloud data obtained from the constant false alarm rate detection and single-pulse angle measurement steps described above is used as input, and the DBSCAN clustering algorithm is used to cluster the point clouds. During the clustering process, the Euclidean distance and velocity differences between point clouds are considered simultaneously. For each cluster, the geometric center of all point clouds within that cluster in the spatial coordinate system is calculated, and this geometric center is used as the final detection location of the corresponding physical target.

[0161] In summary, this application achieves spatial domain clutter suppression through digital beamforming and combines it with Adaptive Virtual Phase Center Alignment (AVPCA) for coherent cancellation of ground clutter in the slow time domain, effectively reducing clutter energy. The introduction of an LMS closed-loop optimization mechanism improves the system's robustness to platform motion errors, resulting in more stable clutter suppression performance compared to the traditional DPCA method. The mask-assisted CFAR employed in this application generates a mask based on a priori ground clutter models, partitions the range-Doppler map, increases the detection threshold in clutter regions, and maintains standard CFAR in background regions, thereby reducing false alarm rates while maintaining detection sensitivity and effectively adapting to non-uniform clutter environments. Simultaneously, beam domain dimensionality reduction significantly reduces data size, and clutter suppression is performed at the front end to reduce subsequent processing pressure. The overall algorithm primarily uses weighted and local operations, exhibiting good parallelism and hardware friendliness, making it suitable for implementation on embedded DSP or FPGA platforms, thus reducing computational complexity.

[0162] The scope of protection for the obstacle detection method based on millimeter-wave radar described in this application is not limited to the order of steps listed in this embodiment. Any solution implemented by adding, subtracting, or replacing steps in the prior art based on the principles of this application is included within the scope of protection of this application.

[0163] This application also provides an obstacle detection system based on millimeter-wave radar. The obstacle detection system based on millimeter-wave radar can implement the obstacle detection method based on millimeter-wave radar described in this application. However, the implementation device of the obstacle detection method based on millimeter-wave radar described in this application includes, but is not limited to, the structure of the obstacle detection system based on millimeter-wave radar listed in this embodiment. All structural modifications and substitutions of the prior art made in accordance with the principles of this application are included within the protection scope of this application.

[0164] like Figure 12 As shown, this embodiment provides an obstacle detection system based on millimeter-wave radar, including: a data acquisition module 21, a data preprocessing module 22, a clutter suppression module 23, a global spectrum analysis module 24, and a point cloud post-processing module 25.

[0165] The data acquisition module 21 is used to acquire the raw sampling data output by each receiving channel of the millimeter-wave radar after analog-to-digital conversion.

[0166] In this embodiment, a frequency-modulated continuous wave signal is transmitted by a millimeter-wave radar, and the echo signal is received. After analog-to-digital conversion, a multi-channel digital intermediate frequency signal is obtained, and the digital intermediate frequency signal is output as the original sampling data.

[0167] The data preprocessing module 22 is used to preprocess the original sampled data to obtain the dimension-reduced distance-slow time-beam data.

[0168] In one embodiment, the original sampled data is compressed in the range direction to obtain a range-dimensional data cube; the range-dimensional data cube is then reduced in dimension by a beam signal to obtain the reduced range-slow-time-beam data.

[0169] The clutter suppression module 23 is used to perform ground clutter suppression on the range-slow time-beam data using an adaptive virtual phase center alignment method to obtain a clutter-suppressed signal.

[0170] In one embodiment, the drone's flight control system fuses data from multiple sensors to obtain the drone's real-time speed; based on the drone's real-time speed, amplitude and phase corrections are performed, and the phase compensation amount of the two-channel data is dynamically adjusted to calculate adaptive weights for virtual phase center alignment; according to different distance units, the adaptive weights are adjusted in segments to generate fine compensation amounts corresponding to each distance unit; the least mean square algorithm is used to adaptively filter the fine compensation amounts to obtain a clutter-suppressed signal.

[0171] The global spectrum analysis module 24 is used to perform velocity-dimensional spectrum analysis and constant false alarm rate detection on the clutter-suppressed signal, and to extract point cloud data of the target obstacle by combining single-pulse angle measurement; the point cloud data includes distance, velocity and angle information.

[0172] In one embodiment, firstly, a velocity-dimensional Fourier transform is performed on the clutter-suppressed signal to construct a range-Doppler map.

[0173] Then, a constant false alarm rate (CFAR) algorithm is used to perform target detection on the range-Doppler map and extract candidate target points. This includes: calculating the clutter ridge region based on the current UAV speed and altitude using the Doppler frequency formula for stationary ground clutter, and generating a mask region on the range-Doppler map; selecting several protection units and reference units on both sides of the unit to be detected, calculating the average value of the reference units on both sides, and multiplying the average value by a coefficient to obtain a basic detection threshold; if the unit to be detected is located within the mask region, a bias constant is superimposed on the basic detection threshold to obtain the final detection threshold; if the unit to be detected is not within the mask region, the basic detection threshold is used as the final detection threshold; comparing the power value of the unit to be detected with the final detection threshold, and outputting the detection value based on the comparison result.

[0174] Furthermore, when the power value of the unit to be detected is greater than the final detection threshold, the unit to be detected is determined to be a candidate target point, and the power value of the unit to be detected is used as the candidate target point; when the power value of the unit to be detected is less than or equal to the final detection threshold, 0 is output, and the unit is determined to be clutter background.

[0175] Finally, single-pulse angle measurement is performed on the candidate target points to obtain the angle information of each candidate target point, and point cloud data containing distance, velocity and angle are obtained.

[0176] The point cloud post-processing module 25 is used to cluster the point cloud data using a spatial clustering algorithm to obtain the spatial coordinates of the target obstacle and finally determine the true location of the target obstacle.

[0177] In one embodiment, a density-based spatial clustering algorithm is used to cluster point clouds, merging point clouds that are spatially and velocally adjacent into a cluster, and calculating the geometric center of each cluster to represent a real physical obstacle.

[0178] The structure and principle of the data acquisition module 21, data preprocessing module 22, clutter suppression module 23, global spectrum analysis module 24, and point cloud postprocessing module 25 correspond one-to-one with the steps in the above-mentioned obstacle detection method based on millimeter-wave radar, so they will not be described again here.

[0179] In summary, the system provided in this application employs a three-transmitter, four-receiver (3T4R) single-chip millimeter-wave radar, combined with digital beamforming technology, to achieve channel data dimensionality reduction at the receiving end. Compared to traditional multi-chip solutions, this invention significantly reduces the number of RF front-ends and data interfaces, lowering system power consumption and size, making it suitable for small UAV platforms. Simultaneously, this application integrates a fuselage state perception module (speed, altitude, etc.) and possesses range segmentation fine-grained compensation capabilities, dynamically adjusting AVPCA weights and mask areas based on current flight parameters. The system maintains a stable detection probability even in non-uniform clutter environments such as near-ground and urban canyons, effectively suppressing false alarms caused by stationary ground objects and low-speed clutter, demonstrating excellent environmental adaptability and dynamic response capabilities.

[0180] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, or methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or modules or units may be electrical, mechanical, or other forms.

[0181] The modules / units described as separate components may or may not be physically separate. The components shown as modules / units may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules / units can be selected to achieve the objectives of the embodiments of this application, depending on actual needs. For example, the functional modules / units in the various embodiments of this application may be integrated into one processing module, or each module / unit may exist physically separately, or two or more modules / units may be integrated into one module / unit.

[0182] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0183] Please see Figure 13 The diagram shows a structural schematic of the electronic device described in an embodiment of this application. Figure 13 As shown, this embodiment provides an electronic device, the electronic device 30 including a memory 32 and a processor 31.

[0184] The memory 32 is used to store computer programs; preferably, the memory 32 includes various media that can store program code, such as ROM, RAM, magnetic disk, USB flash drive, memory card or optical disk.

[0185] Specifically, memory 32 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. Electronic device 30 may further include other removable / non-removable, volatile / non-volatile computer system storage media. Memory 32 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application. It is understood that memory 32 may be volatile memory or non-volatile memory, or both. Non-volatile memory may be read-only memory (ROM) or programmable read-only memory (PROM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable categories of memory.

[0186] The processor 31 is connected to the memory 32 and is used to execute the computer program stored in the memory 32 so that the electronic device 30 performs the obstacle detection method based on millimeter-wave radar as described in any embodiment of this application.

[0187] Optionally, the processor 31 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0188] This application also provides a computer-readable storage medium. Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a computer-readable storage medium, which is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof. The storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state drive (SSD)).

[0189] This application embodiment may also provide a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application embodiment are generated. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0190] When the computer program product is executed by a computer, the computer performs the method described in the foregoing method embodiments. The computer program product can be a software installation package; when the foregoing method is required, the computer program product can be downloaded and executed on the computer.

[0191] In summary, the obstacle detection method, system, and device based on millimeter-wave radar provided in this application have the following beneficial effects:

[0192] This application provides an obstacle detection method, system, and device based on millimeter-wave radar. It achieves efficient suppression of ground clutter under platform motion and attitude disturbances through adaptive virtual phase center alignment (AVPCA) and LMS closed-loop filtering. Combined with a mask-assisted CA-CFAR detection strategy, it reduces false alarms while maintaining sensitivity to slow-moving targets. Furthermore, it employs digital beamforming dimensionality reduction and piecewise linear approximation to significantly reduce computational complexity. Coupled with DBSCAN clustering and a single-chip 3T4R hardware architecture, the system possesses robust, real-time, and low-power obstacle detection capabilities in low-altitude, highly cluttered environments, making it particularly suitable for embedded deployment on small UAV platforms. Moreover, this application has broad applicability. Therefore, this application effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0193] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0194] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. An obstacle detection method based on millimeter-wave radar, characterized in that, The method includes: Acquire the raw sampling data output from each receiving channel of the millimeter-wave radar after analog-to-digital conversion; The original sampled data is preprocessed to obtain the dimension-reduced distance-slow-time-beam data; An adaptive virtual phase center alignment method is used to suppress ground clutter in the range-slow-time-beam data to obtain the clutter-suppressed signal. The clutter-suppressed signal is subjected to velocity-dimensional spectrum analysis and constant false alarm rate detection, and point cloud data of the target obstacle is extracted by combining single-pulse angle measurement; the point cloud data includes distance, velocity and angle information; A spatial clustering algorithm is used to cluster the point cloud data to obtain the spatial coordinates of the target obstacle, and finally determine the true location of the target obstacle.

2. The obstacle detection method based on millimeter-wave radar according to claim 1, characterized in that, The original sampled data is preprocessed to obtain the dimensionality-reduced range-slow-time-beam data, including: The original sampled data is compressed in the distance direction to obtain a data cube in the distance dimension; The distance-dimensional data cube is reduced in dimensionality by using beam signals to obtain the reduced distance-slow-time-beam data.

3. The obstacle detection method based on millimeter-wave radar according to claim 1, characterized in that, Ground clutter suppression is performed on the range-slow-time-beam data using an adaptive virtual phase center alignment method, resulting in a clutter-suppressed signal including: The drone's real-time speed is obtained by integrating data from multiple sensors through the drone's flight control system. Based on the real-time speed of the UAV, amplitude and phase corrections are performed, the phase compensation amount of the two-channel data is dynamically adjusted, and adaptive weights for virtual phase center alignment are calculated. Based on the different distance units, the adaptive weights are adjusted in segments to generate fine compensation amounts corresponding to each distance unit. The least mean square algorithm is used to adaptively filter the fine compensation amount to obtain the signal after clutter suppression.

4. The obstacle detection method based on millimeter-wave radar according to claim 3, characterized in that, Multiple sensors include: an inertial detection unit and a global navigation satellite system; the real-time speed of the UAV is obtained by fusing data from multiple sensors through the UAV's flight control system, including: The first velocity component is calculated by integrating time using the accelerometer inside the inertial detection unit. The second velocity component is calculated using the global navigation satellite system. The first velocity component and the second velocity component are fused to obtain the real-time velocity of the UAV.

5. The obstacle detection method based on millimeter-wave radar according to claim 3, characterized in that, The formula for correcting the adaptive weights is: in, Represents the complex weighting factor; Indicates the amplitude scaling factor; Indicates compensation for phase difference; This represents the complex value of the i-th distance gate in the n-th linear frequency modulated signal of the coherent source used as a reference. This represents the secondary channel of the compensation signal; Indicates wavelength; Indicates the downward angle; Indicates the pulse repetition period; This represents the difference between the ideal displacement and the actual displacement of the airborne radar during its movement. Indicates the speed of radar movement; The formula for the segmented adjustment includes: The formula for segmentation based on the relationship between H and R is: The weight update formula in the adaptive filtering is: in, Indicates the iteration step size; This indicates the elimination of residuals; This represents the iterative weight parameters.

6. The obstacle detection method based on millimeter-wave radar according to claim 1, characterized in that, The clutter-suppressed signal undergoes velocity-dimensional spectrum analysis and constant false alarm rate detection, and point cloud data of the target obstacle is extracted using monopulse angle measurement, including: Based on the clutter-suppressed signal, a velocity-dimensional Fourier transform is performed to construct a range-Doppler spectrum; The constant false alarm rate algorithm is used to perform target detection on the range-Doppler spectrum and extract candidate target points; Single-pulse angle measurement is performed on the candidate target points to obtain the angle information of each candidate target point, and point cloud data containing distance, velocity and angle are obtained.

7. The obstacle detection method based on millimeter-wave radar according to claim 6, characterized in that, The constant false alarm rate (CFAR) algorithm is used to perform target detection on the range-Doppler spectrum, and candidate target points are extracted, including: Based on the current speed and altitude of the UAV, the region where the clutter ridge is located is calculated using the Doppler frequency formula of the stationary ground clutter, and a mask region is generated on the range-Doppler spectrum. Several protection units and reference units are selected on both sides of the unit to be tested. The average value of the reference units on both sides is calculated. The average value is multiplied by a coefficient to obtain the basic detection threshold. If the unit to be detected is located within the mask area, a bias constant is superimposed on the basic detection threshold to obtain the final detection threshold; if the unit to be detected is not located within the mask area, the basic detection threshold is used as the final detection threshold. The power value of the unit to be detected is compared with the final detection threshold, and the detection value is output based on the comparison result; include: When the power value of the unit to be detected is greater than the final detection threshold, the unit to be detected is determined to be a candidate target point, and the power value of the unit to be detected is used as the candidate target point. When the power value of the unit to be detected is less than or equal to the final detection threshold, the output is 0, and the unit is determined to be clutter background.

8. The obstacle detection method based on millimeter-wave radar according to claim 1, characterized in that, Spatial clustering algorithms are used to cluster the point cloud data to obtain the spatial location coordinates of the target obstacle, including: A density-based spatial clustering algorithm is used to cluster point clouds, merging point clouds that are spatially and velocally adjacent into a cluster, and calculating the geometric center of each cluster to represent a real physical obstacle.

9. An obstacle detection system based on millimeter-wave radar, characterized in that, The system includes: The data acquisition module is used to acquire the raw sampled data output from each receiving channel of the millimeter-wave radar after analog-to-digital conversion; The data preprocessing module is used to preprocess the original sampled data to obtain the dimension-reduced distance-slow-time-beam data; The clutter suppression module is used to perform ground clutter suppression on the range-slow time-beam data using an adaptive virtual phase center alignment method to obtain the clutter-suppressed signal. The global spectrum analysis module is used to perform velocity-dimensional spectrum analysis and constant false alarm rate detection on the clutter-suppressed signal, and to extract point cloud data of the target obstacle by combining it with single-pulse angle measurement; the point cloud data includes: distance, velocity and angle information; The point cloud post-processing module is used to cluster the point cloud data using a spatial clustering algorithm to obtain the spatial coordinates of the target obstacle and ultimately determine the true location of the target obstacle.

10. An electronic device, characterized in that, include: Processor and memory; The memory is used to store computer programs; The processor is connected to the memory and is used to execute the computer program stored in the memory to enable the electronic device to perform the obstacle detection method based on millimeter-wave radar as described in any one of claims 1 to 8.

Images