Object detection method and device
By converting multi-base radar into single-base virtual radar, the existing packaging detection methods are solved, and high-precision and high-efficiency object detection are achieved, reducing costs.
Patent Information
- Application Number
- CN202210950421.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-09
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2042-08-09
AI Technical Summary
The existing product packaging inspection methods have low accuracy, low efficiency and high cost, making it difficult to meet the needs of production automation.
The multi-base radar is converted into a single-base virtual radar, and by determining the phase and frequency of the preset signal, the target to be tested is detected, the signal parameters are obtained, the received signal is corrected, the target image is generated, and the image training detection model is used for identification.
It improves the accuracy and efficiency of object detection, reduces costs, and achieves high-precision and high-efficiency packaging inspection.
Smart Images

Figure CN115453478B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of signal processing, and in particular relates to an object detection method and device. Background Art
[0002] The digital transformation of manufacturing companies is unstoppable. Equipment with intelligent and automated functions will become the mainstream equipment on production lines, and the market prospects are huge.
[0003] With the continuous increase of labor costs, production automation has become an inevitable trend. As the last link in the industrial value chain, packaging has received more and more attention from manufacturers. Research has found that the current product packaging testing methods on the market have low accuracy, low efficiency and high cost. Summary of the invention
[0004] The present invention provides an object detection method and device, which has the characteristics of high precision, high efficiency and low cost.
[0005] In a first aspect, the present application provides an object detection method, comprising: converting a multi-base radar into a single-base virtual radar, and determining a first parameter of a preset signal based on the single-base virtual radar, wherein the preset signal is a signal reflected by a corner reflector, and the first parameter includes a phase and a frequency of the preset signal; detecting the target to be measured to obtain a second parameter of the signal of the target to be measured, and the first parameter includes a phase and a frequency of the signal of the target to be measured; determining a target position of the target to be measured based on the first parameter and the second parameter; correcting a measurement signal of a receiving antenna based on the target position to obtain a corrected signal; and obtaining an image of the target to be measured based on the corrected signal.
[0006] The step of determining the first parameter of a preset signal based on the single-base virtual radar includes: determining the preset signal using the single-base virtual radar; and performing Fourier transform processing on the signal reflected by the corner reflector to obtain the first parameter of the preset signal.
[0007] Among them, the step of detecting the target to be measured and obtaining the second parameter of the signal of the target to be measured includes: detecting the target to be measured to obtain a first echo signal; sampling the first echo signal based on the time of the first echo signal to obtain multiple first beat signals; performing Fourier transform processing on the first beat signal to determine the phase and frequency of the first beat signal; determining the second parameter based on the phase and frequency of the first beat signal and the phase and frequency of the measurement signal of the receiving antenna.
[0008] Among them, the step of determining the target position of the target to be measured based on the first parameter and the second parameter includes: determining the phase difference between the phase in the first parameter and the phase in the second parameter; and determining the target position of the target to be measured based on the phase difference by using least squares curve fitting.
[0009] The step of correcting the measurement signal of the receiving antenna based on the target position to obtain a corrected signal includes: constructing a reference beat signal model based on the target position; and processing the measurement signal using the reference beat signal model to obtain a corrected signal.
[0010] Among them, the step of obtaining the image of the target to be measured based on the corrected signal includes: performing Fourier transform in azimuth and altitude dimensions on the corrected signal to obtain a first signal; performing phase compensation on the first signal to obtain a third signal; and performing inverse Fourier transform in azimuth and altitude dimensions on the third signal to obtain the image of the target to be measured.
[0011] The method further comprises: training a target detection model using an image of the target to be detected, wherein the image of the target to be detected is annotated with a category of the target to be detected.
[0012] The method further includes: using the target detection model to identify the image of the target to be detected, and determining the category of the target to be detected.
[0013] In a second aspect, the present application provides an object detection device, comprising: a processing module, used to convert a multi-base radar into a single-base virtual radar, and determine a first parameter of a preset signal based on the single-base virtual radar, wherein the preset signal is a signal reflected by a corner reflector, and the first parameter includes a phase and a frequency of the preset signal; a detection module, used to detect the target to be measured and obtain a second parameter of the signal of the target to be measured, wherein the first parameter includes a phase and a frequency of the signal of the target to be measured; a position determination module, used to determine a target position of the target to be measured based on the first parameter and the second parameter; a correction module, used to correct the measurement signal of the receiving antenna based on the target position to obtain a corrected signal; and an imaging module, used to obtain an image of the target to be measured based on the corrected signal.
[0014] The beneficial effects of the present invention are different from those of the prior art. The object detection method of the present invention converts a multi-base radar into a single-base virtual radar, and determines the first parameter of a preset signal based on the single-base virtual radar, wherein the preset signal is a signal reflected by a corner reflector, and the first parameter includes the phase and frequency of the preset signal; detects the target to be measured to obtain the second parameter of the signal of the target to be measured, and the first parameter includes the phase and frequency of the signal of the target to be measured; determines the target position of the target to be measured based on the first parameter and the second parameter; corrects the measurement signal of the receiving antenna based on the target position to obtain a corrected signal; and obtains an image of the target to be measured based on the corrected signal. The detection accuracy of the image can be improved, and the efficiency is high and the cost is low. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 is a flow chart of a first embodiment of an object detection method of the present invention;
[0016] Figure 2 is a flow chart of a second embodiment of the object detection method of the present invention;
[0017] Figure 3 is a flow chart of a third embodiment of an object detection method of the present invention;
[0018] Figure 4 FIG. 4 is a schematic structural diagram of a first embodiment of an object detection device according to the present invention. DETAILED DESCRIPTION
[0019] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods. The aforementioned and other technical contents, features and effects of the present invention can be clearly presented in the following detailed description of the specific implementation methods in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. The attached drawings are only for reference and explanation purposes and are not used to limit the technical solutions of the present invention.
[0020] See also Figure 1 , Figure 1 The flowchart of an embodiment of the object detection method of the present invention specifically includes:
[0021] Step S11: converting the multistatic radar into a monostatic virtual radar, and determining a first parameter of a preset signal based on the monostatic virtual radar, wherein the preset signal is a signal reflected by a corner reflector, and the first parameter includes a phase and a frequency of the preset signal.
[0022] In the embodiment of the invention, the radar can also be represented as an antenna array, which includes multiple receiving antennas and transmitting antennas. The conversion of a multi-base radar to a mono-base virtual radar can be called the conversion of a multi-base antenna array to a mono-base antenna array.
[0023] In a multistatic array, the transmitting and receiving antennas are not in the same location, and the multiple-input multiple-output (MIMO) array operates sequentially or simultaneously. The multistatic array topology in the microwave imaging system is converted into a monostatic array using the effective phase center principle.
[0024] The conversion method is as follows. The coordinates of the effective phase center antenna in the xy plane can be expressed as:
[0025]
[0026]
[0027] Where (x c ,y c ) represents the coordinates of the effective phase center antenna, (x T ,y T ) and (x R ,y R ) are the coordinates of the transmitting antenna and the receiving antenna respectively.
[0028] The reference point (x ref ,y ref ,z ref ) is defined as the transformation of calibrated multi-static to mono-static. For the echo data set s(x T ,y T ,x R ,y R ,k), The transformed data set is given by:
[0029]
[0030] where s ref is the echo data set that would be received by an effective monostatic array antenna imaging a single point scatterer if the target domain contained only point scatterers at the reference point in a multistatic array. And:
[0031]
[0032] Where R T (R R ) is the distance between the transmitting (receiving) antenna and the single point scatterer in the multi-base array, R C is the distance between the antenna and the single point scatterer in an assumed monostatic array.
[0033] Specifically, the multi-base radar can be converted into a single-base virtual radar through the above formulas (9), (10), and (11).
[0034] Further determine the first parameter of the preset signal based on the single-base virtual radar. In one embodiment, the preset signal is determined using the single-base virtual radar; the signal reflected by the corner reflector is Fourier transformed to obtain the first parameter of the preset signal. It should be noted that the preset signal is a signal reflected by the corner reflector, and the first parameter includes the phase and frequency of the preset signal. Specifically, the corner reflector is a reference point. First, accurately estimate the position of the reference target, that is, the corner reflector, and use the single-base virtual radar to collect data along the vertical axis and the horizontal axis as a reference channel.
[0035] The total round-trip delay of the backscattered data is The frequency and phase of the beat signal measured in the is directly related. The captured broadband beat signal is defined as s(x′,y′,t) as a three-dimensional function of the measurement points in the time and xy domains. The beat frequency and phase of each measurement point are needed to estimate the accurate range profile of the target. The three-dimensional beat signal model reflected from an ideal target point (corner reflector) is defined as:
[0036]
[0037] Among them, f b (x′, y′)=Kτ(x′, y′), φ(x′, y′)=2πf0τ(x′, y′) are the frequency and phase of the beat signal, respectively, and are functions of the round-trip delay. Assume uniform time sampling, that is, s[x′,y′,n]=s(x′,y′,nT s ), where T s is the sampling period and h=[0,1,…N-1]. Then the frequency and phase of the sampled beat signal are estimated by performing N-point discrete time Fourier transform on it:
[0038]
[0039] Among them, w b (x′,y′)=2πf b (x′, y′) is the beat frequency of the corner reflector. Since the beat signal is obtained by uniform time sampling, the above formula can be implemented by time FFT.
[0040] Specifically, the first parameter of the preset signal can be obtained by the above formula (1) and formula (2).
[0041] Step S12: Detect the target to be measured to obtain a second parameter of the signal of the target to be measured, wherein the first parameter includes a phase and a frequency of the signal of the target to be measured.
[0042] Specifically, the target to be measured is detected to obtain a first echo signal; the first echo signal is sampled based on the time of the first echo signal to obtain multiple first beat signals; the first beat signal is Fourier transformed to determine the phase and frequency of the first beat signal; the second parameter is determined based on the phase and frequency of the first beat signal and the phase and frequency of the measurement signal of the receiving antenna.
[0043] Specifically, the target distance varies very little within a single beat signal. Therefore, in order to obtain a more accurate distance distribution, which is a function of the round-trip delay in equation (4), we need to use the phase of the beat signal φ(x',y') in equation (3):
[0044]
[0045]
[0046] For all measurement points, if you select Then the beat frequency corresponds to the coordinate of the FFT output peak, and the complex value at this coordinate contains only the phase term e jφ(x′,y′) . Ignore the phase residue caused by the finite FFT.
[0047] Step S13: determining the target position of the target to be measured based on the first parameter and the second parameter.
[0048] Determine a phase difference between the phase in the first parameter and the phase in the second parameter; and determine a target position of the target to be measured based on the phase difference by using a least squares curve fitting method.
[0049] Specifically, the position (x, y, z) of a point target with reference to the scanning geometry can be estimated by applying a least squares curve fitting method to the measured beat signal phase. These methods lead directly to an estimate of the target position by finding a set of parameters that minimizes the squared error between the modeled phase and the measured phase as:
[0050]
[0051] Here, we assume that the actual distance between the radar aperture and the point target shown in the range FFT output plot is available in the measured beat signal phase. However, in the actual environment of an FMCW signal scheme, the range FFT operation can be used to provide a target range estimate. Where R(xr ,y′)=cτ(x′,y′) / 2 is the actual distance from the target to the known measurement point (x′,y′), R b is the distance deviation caused by hardware defects. Therefore, the last step in the position estimation problem is to compensate for the distance deviation R of the reference channel. b , that is, step S14.
[0052] Step S14: Correcting the measurement signal of the receiving antenna based on the target position to obtain a corrected signal.
[0053] Specifically, a reference beat signal model is constructed based on the target position; and the measurement signal is processed using the reference beat signal model to obtain a corrected signal.
[0054] The distance deviation is estimated by using the coupling between the transmitting and receiving antenna elements. By observing the distance distribution of the beat signal, the target-independent distance deviation caused by mutual coupling can be obtained. Let us assume that the reference channel consists of u ∈R 3 and r v ∈R 3 The u-th transmitter and v-th receiver element of . Then the Euclidean distance between the corresponding transceiver antenna pairs can be obtained using |r u -r v |Modeled reference beat signal to estimate distance deviation.
[0055] In one embodiment, the estimated target location (x, y, z) is used to create a reference backscatter signal model for each channel of the MIMO array. First, the total round-trip delay of the FMCW signal reflected from the target point by the u-th transmitting antenna and the v-th receiving antenna is defined as And the corresponding transmission gain a i We model the delay between the antenna pair as a superposition of the common instrument delay and the residual delay between the antenna elements: Ignoring additive noise and RVP terms, the uncalibrated measured beat signal can be defined as:
[0056]
[0057] in, is the reference beat signal model, f i =Kτ i is the beat frequency that causes the system to have distance deviation, η l is the residual complex gain factor. Given the measured signal The corrected error signal can be calculated by a simple demodulation process:
[0058]
[0059] The * sign indicates the complex conjugate operation. From (13), we can estimate f i and η l This boils down to the problem of parameter estimation of a single-frequency complex tone obtained from noisy observations:
[0060]
[0061] in Defined as:
[0062]
[0063] If the data If the sample is uniformly spaced in t, then FFT can be used to obtain The complex gain factor of each transceiver is η l The estimated value can be Substitute into (12) to calculate.
[0064] Step S15: obtaining an image of the target to be measured based on the corrected signal.
[0065] After the correction in step S14, the corrected signal is expressed as:
[0066]
[0067] Where (x′, y′) represents the position of the measurement point, (x, y, z) represents the position of the target pixel, R represents the distance between the measurement point and the target pixel, and p(x, y, z) represents the reflectivity of the pixel. Represents the amplitude attenuation factor.
[0068] The corrected signal is subjected to a Fourier transform in the azimuth dimension and the altitude dimension to obtain a first signal. Specifically, the echo data is subjected to a two-dimensional Fourier transform in the azimuth dimension and the altitude dimension, that is, (12) is subjected to a two-dimensional Fourier transform to obtain:
[0069]
[0070] Using the stationary phase method, Solving it, we can get:
[0071]
[0072] Where z=z0 is the plane where the measuring point is located. Substituting into (14), we can obtain:
[0073]
[0074] Phase compensation is performed on the first signal to obtain a third signal. Specifically, phase compensation is performed on the data after two-dimensional FFT, and the compensation factor is get in,
[0075]
[0076] The integral part of the above formula is consistent with the three-dimensional Fourier transform. However, k z It is not evenly spaced, which does not meet the requirements of Fourier transform. We need to process it at equal intervals, and we can perform interpolation processing on it. This system only needs to complete two-dimensional imaging, so we can treat the target as a plane target. z takes a fixed value, then equation (16) changes to
[0077]
[0078] The integral part of the above formula is in the form of two-dimensional Fourier transform.
[0079] The third signal is subjected to an inverse Fourier transform in the azimuth dimension and the height dimension to obtain an image of the target to be measured. Specifically, a two-dimensional IFFT transform is simultaneously performed on the left and right sides of (17) to obtain a planar image of the target.
[0080]
[0081] The data is reconstructed into a two-dimensional image on the computing platform AI-BoxX. Finally, the empty package in the box is determined by imaging based on the reconstructed image. At the same time, AI-BoxX is also responsible for configuring the millimeter-wave radar front end and controlling the stepper motor.
[0082] In the present application, a millimeter-wave array radar system is used for object detection. Specifically, a radar imaging subsystem is composed of a 77GHz highly integrated millimeter-wave sensor, a one-dimensional linear array antenna, and an FPGA data acquisition module. The one-dimensional linear array antenna scans the vertical plane by electronic scanning, and the horizontal plane is obtained by the horizontal movement of the object under test. The radar sends and receives FMCW signals at a rate of 20ms per frame.
[0083] In the radar imaging process, the radar signal illuminates the target on a planar aperture. The reflected signal is then recorded by the receiver at different points in space, and finally the reflected amplitude focused image of the target is mathematically formed using the coherently recorded sampled data.
[0084] The radar wavelength is 3.896mm, and the time interval of each frame is about 1 / 40 = 0.025s, that is, 25ms. Each movement is about 220mm, and the number of valid frames in the acquisition is about 220 / 1 = 220. After completing the above system parameter configuration, enable the Cascade_Configuration_MIMO.lua script to complete the system parameter configuration, and perform the following steps to complete the data acquisition:
[0085] (1). The radar starts data acquisition (the data acquisition script Cascade_Capture.lua is called by MATLAB to start data acquisition);
[0086] (2). The target coordinates start to move in one direction at 2400 mm / min (controlled by calling the serial port through MATLAB);
[0087] (3) Collect 240 frames in total, discard the first 10 frames and the last 10 frames, and use the middle 220 frames.
[0088] ADC is used for data transmission, the collected radar data is parsed, and then transmitted to the computing platform. Specifically, the FPGA data acquisition module uses LVDS to deserialize and synchronize the collected serial signals. At the same time, a PCB board for transferring the LVDS high-speed serial interface is drawn and manufactured, and it is connected to the Xilinx ZYNQ SoC and Intel Cyclone V SoC. The logic code for data transmission is written in Verilog and C. The final data stream is transmitted to the 32-bit parallel port via LVDS.
[0089] The IWR2243EVM circuit board provides a Samtec 60-pin high-density connector, and a three-seat transfer PCB board is drawn and manufactured. The FPGA board and the IWR2243 EVM circuit board are connected at the same time, and the impedance is controlled at about 50 ohms to achieve LVDS impedance matching.
[0090] Ethernet data transmission. The intermediate frequency sampling data of the four-channel radar ADC is obtained after deserialization and synchronization by the Xilinx ZYNQ SoC, and is transmitted to the Intel Cyclone V SoC using a 32-bit low-speed parallel bus with 1 clock line + 32 data lines. After the FPGA logic captures the data on the parallel bus, it packages the data and sends it to the ARM processor side through the DMA controller. The LwIP protocol stack runs in bare metal mode on the ARM processor, and finally the data is packetized and sent from Ethernet to the AI-BoxX through the UDP protocol.
[0091] The imaging results are obtained on the AI-BoxX platform using the Inverse Synthetic Aperture Radar (ISAR) processing technology.
[0092] Read the imaging results on the AI-BoxX platform. Specifically, read the bin file collected at the time from the folder. The data acquisition method mainly adopts the method in the demo, but the demo processes by frame, so you need to rewrite the top-level function of data reading and directly organize the complete data. The data format of a frame read by the demo is: numSamplePerChirp* numChirpPerLoop*numRXPerDevice*numChirpPerLoop (the data shape should be 256*1*16*12 under the current configuration). By drawing the RX and TX and the virtual antenna array positions, find the corresponding relationship between Tx, Rx and the corresponding virtual array, and provide a basis for subsequent data processing and correction.
[0093] During the imaging process, the near-field moving target is modeled in the azimuth dimension, and its echo signal has a synthetic aperture form in the azimuth dimension; at the same time, the array echo signal model is established in the height dimension using the MIMO linear array combined with the equivalent phase center principle. In addition, under near-field conditions, the echo data is firstly subjected to range dimension pulse compression, and then the data is rearranged by azimuth interpolation to achieve distance correction, and finally the height dimension is subjected to range migration correction to eliminate the data error caused by the near field; after obtaining the distance-corrected data, the echo signal is subjected to refined motion compensation to eliminate the refined phase error caused by the target motion. The specific correction process is as follows: Figure 1 shown.
[0094] See also Figure 2 , Figure 2 This is a flow chart of a second embodiment of the object detection method of the present application, which is similar to the above Figure 1 Compared with the first embodiment shown in the figure, the difference is that this embodiment also includes:
[0095] Step S16: training a target detection model using the image of the target to be detected, wherein the image of the target to be detected is annotated with the category of the target to be detected.
[0096] Specifically, collect images of the target to be tested to construct a millimeter wave radar ISAR image training set. Through the detection package, continuously obtain imaging results to form an image data set that can be used for network training. Specifically, use the constructed training set to complete the YOLO network training. Specifically, YOLO target detection and recognition has the following features:
[0097] ①YOLO structure. After removing the candidate region step, YOLO's structure is very simple, which is a simple convolution, pooling and two layers of full connection. The biggest difference is that the final output layer uses a linear function as the activation function, because it needs to predict the position of the bounding box (numerical type), not just the probability of the object.
[0098] ② Input and output. In the YOLO network, the convolutional layer is connected to two fully connected layers at the end. The fully connected layer requires a vector of a fixed size as input, so the original image is required to have a fixed size.
[0099] ③ Training samples. As a supervised learning, we need to construct training samples first so that the model can learn from them.
[0100] ④ Loss function. The loss is the deviation between the actual output value of the network and the sample label value.
[0101] ⑤ Training. YOLO first uses the ImageNet dataset to pre-train the first 20 layers of the convolutional network, and then uses the complete network to train and predict object recognition and positioning on the PASCAL VOC dataset.
[0102] ⑥ Prediction. YOLO uses the NMS (Non-maximal suppression) algorithm to extract the most likely objects and locations.
[0103] ⑦ Non-maximum suppression. The core idea of NMS is to select the one with the highest score as the output, remove those that overlap with the output, and repeat this process until all candidates have been processed.
[0104] Based on the YOLO target detection network features and image training set, the mean square error is used as the loss function to optimize the YOLO model parameters, that is, the mean square error between the S*S*(B*5+C)-dimensional vector output by the network and the corresponding S*S*(B*5+C)-dimensional vector of the real image, where S*S is the number of image grids, C is the number of target categories, and B is the number of detection boxes corresponding to each network; the image to be detected is divided into S*S grids. Next, for each grid, the YOLO network is used to predict the position of the detection box and the category of the target in the detection box (empty or non-empty, deformed or non-deformed).
[0105] See also Figure 3 , Figure 3 FIG. 1 is a flow chart of a third embodiment of the object detection method of the present invention, which is similar to the above Figure 2 The difference is that this embodiment further includes:
[0106] Step S17: using the target detection model to identify the image of the target to be detected, and determining the category of the target to be detected.
[0107] The intelligent target detection software and YOLO network are run on the computing platform AI-BoxX. The generated radar imaging image is used as input, and the network is directly used to predict the location of the detection box and the category of the target in the detection box. The classification and recognition results are displayed through a graphical user interface (GUI) and the screen.
[0108] The effects of the present invention are further verified by the following specific system parameter settings and index analysis:
[0109] 1. System and scene parameter settings:
[0110] Table 1 System and scene parameters
[0111]
[0112] 2. System indicator verification
[0113] 2.1 Radar / RF Part
[0114] (1) Frequency band, bandwidth: E band, 77-81 GHz, bandwidth 4 GHz
[0115] (2) Radar transmit power:
[0116] According to the radar equation, the relationship between the range and the transmission power is:
[0117]
[0118] It can be calculated that when the effective distance is R = 0.4m, the transmission power should be P t =36.7mW (for 9dB noise figure), P t =9.2mW (for 3dB noise figure).
[0119] Maximum transmit power density Meet the national standard of human body radiation 100mW / ㎡. Note: The noise coefficient can be adjusted according to the actual low noise amplifier, and the pulse width can also be reduced to speed up the imaging speed (the pulse width of the L3 system is only 5μs).
[0120] (3) Beam width: 53° (left is directional antenna) / 90° (right is horizontal directional antenna):
[0121]
[0122] (4) PRF / blur distance / expected resolution and other performance:
[0123] Assume that the horizontal distance between the antenna and the target is R = 2m, and the maximum rotation radius of the linear array is r = 1m, then the instantaneous slant distance between the antenna and the target is:
[0124]
[0125] Where, ω is the angular velocity of the linear array, which can be taken as d is the vertical distance between the target point and the center of the linear array, and its value range is 0≤d≤1m.
[0126]
[0127] The maximum and minimum values obtained by Matlab are 5.62 and -5.62 respectively.
[0128] Doppler shift
[0129] Doppler bandwidth PRF=8000Hz
[0130] Unambiguous distance
[0131] Track plane resolution: (Resolution can reach 0.005m during simulation analysis)
[0132] Distance resolution:
[0133] Millimeter-wave radar meets the system requirements of the empty packet detection system in terms of parameter configuration and indicators.
[0134] See also Figure 4 , is a schematic diagram of the structure of an embodiment of the object detection device of the present invention, which specifically includes a processing module 41, a detection module 42, a position determination module 43, a correction module 44, and an imaging module.
[0135] Among them, the processing module 41 is used to convert the multi-base radar into a single-base virtual radar, and determine the first parameter of the preset signal based on the single-base virtual radar, the preset signal is the signal reflected by the corner reflector, and the first parameter includes the phase and frequency of the preset signal. The detection module 42 is used to detect the target to be measured and obtain the second parameter of the signal of the target to be measured, and the first parameter includes the phase and frequency of the signal of the target to be measured. The position determination module 43 is used to determine the target position of the target to be measured based on the first parameter and the second parameter. The correction module 44 is used to correct the measurement signal of the receiving antenna based on the target position to obtain the corrected signal. The imaging module 45 is used to obtain the image of the target to be measured based on the corrected signal.
[0136] The above is only an implementation method of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. An object detection method, characterized in that: include: The multi-base radar is converted into a single-base virtual radar by using the effective phase center principle, and a first parameter of a preset signal is determined based on the single-base virtual radar, wherein the preset signal is a signal reflected by a corner reflector, and the first parameter includes a phase and a frequency of the preset signal; Detecting a target to be measured to obtain a second parameter of a signal of the target to be measured, wherein the second parameter includes a phase and a frequency of the signal of the target to be measured; Determine the target position of the target to be measured based on the first parameter and the second parameter; Correcting the measurement signal of the receiving antenna based on the target position to obtain a corrected signal; Based on the corrected signal, an image of the target to be measured is obtained; Training a target detection model using an image of the target to be detected, wherein the image of the target to be detected is annotated with the category of the target to be detected; Using the target detection model to identify the image of the target to be detected, and determine the category of the target to be detected; Wherein, the step of detecting the target to be measured and obtaining the second parameter of the signal of the target to be measured includes: detecting the target to be measured and obtaining a first echo signal; sampling the first echo signal based on the time of the first echo signal to obtain multiple first beat signals; performing Fourier transform processing on the first beat signal to determine the phase and frequency of the first beat signal; determining the second parameter based on the phase and frequency of the first beat signal and the phase and frequency of the measurement signal of the receiving antenna; Wherein, the step of determining the target position of the target to be measured based on the first parameter and the second parameter includes: determining the phase difference between the phase in the first parameter and the phase in the second parameter; determining the target position of the target to be measured based on the phase difference by using a least squares curve fitting method; The step of correcting the measurement signal of the receiving antenna based on the target position to obtain a corrected signal includes: constructing a reference beat signal model based on the target position; processing the measurement signal using the reference beat signal model to obtain a corrected signal; Among them, the step of obtaining the image of the target to be measured based on the corrected signal includes: performing Fourier transform in azimuth and altitude dimensions on the corrected signal to obtain a first signal; performing phase compensation on the first signal to obtain a third signal; and performing inverse Fourier transform in azimuth and altitude dimensions on the third signal to obtain the image of the target to be measured.
2. The method according to claim 1, characterized in that The step of determining the first parameter of the preset signal based on the single-base virtual radar comprises: Determining the preset signal using the single-base virtual radar; Perform Fourier transform processing on the signal reflected by the corner reflector to obtain the first parameter of the preset signal.
3. An object detection device, characterized in that: include: A processing module, which uses an effective phase center principle to convert a multi-base radar into a single-base virtual radar, and determines a first parameter of a preset signal based on the single-base virtual radar, wherein the preset signal is a signal reflected by a corner reflector, and the first parameter includes a phase and a frequency of the preset signal; A detection module, used to detect the target to be detected and obtain a second parameter of the signal of the target to be detected, wherein the second parameter includes a phase and a frequency of the signal of the target to be detected; A position determination module, used to determine the target position of the target to be measured based on the first parameter and the second parameter; A correction module, used to correct the measurement signal of the receiving antenna based on the target position to obtain a corrected signal; An imaging module, used for obtaining an image of the target to be measured based on the corrected signal; A recognition module is used to train a target detection model using an image of the target to be detected, wherein the image of the target to be detected is annotated with the category of the target to be detected; and to recognize the image of the target to be detected using the target detection model to determine the category of the target to be detected; The detection module is specifically used to: detect the target to be measured to obtain a first echo signal; sample the first echo signal based on the time of the first echo signal to obtain multiple first beat signals; perform Fourier transform processing on the first beat signal to determine the phase and frequency of the first beat signal; determine the second parameter based on the phase and frequency of the first beat signal and the phase and frequency of the measurement signal of the receiving antenna; The position determination module is specifically used to: determine the phase difference between the phase in the first parameter and the phase in the second parameter; and determine the target position of the target to be measured based on the phase difference by using a least squares curve fitting method; The correction module is specifically used to: construct a reference beat signal model based on the target position; process the measurement signal using the reference beat signal model to obtain a corrected signal; Among them, the imaging module is specifically used to: perform Fourier transform in azimuth and altitude dimensions on the corrected signal to obtain a first signal; perform phase compensation on the first signal to obtain a third signal; perform inverse Fourier transform in azimuth and altitude dimensions on the third signal to obtain an image of the target to be measured.
