Near-field MIMO radar data processing method based on multistage background suppression and target enhancement

Abstract

The embodiment of the invention provides a near-field MIMO radar data processing method based on multistage background suppression and target enhancement, and is applied to the technical field of radar signal processing. The method comprises the following steps: acquiring a real-time background template and a real-time image frame of a target scene; calculating to obtain differential image data corresponding to the real-time image frame based on the real-time background template and the real-time image frame; performing Gaussian filtering processing on the difference image data to obtain Gaussian filtering image data; performing morphological processing on the Gaussian filtering image data to obtain morphological image data; performing target enhancement processing on the morphological image data to obtain target enhanced image data corresponding to the real-time image frame; and outputting the target enhanced image data. The technical effect of improving the radar imaging precision is achieved.

Description

Technical Field

[0001] This application relates to the field of radar signal processing technology, and in particular to a near-field MIMO radar data processing method based on multi-level background suppression and target enhancement. Background Technology

[0002] In scenarios such as security inspection, safety protection, and industrial inspection, where there are stringent requirements for high-precision target identification and background suppression, radar needs to quickly identify small targets in complex dynamic environments while suppressing static and dynamic clutter.

[0003] In existing technologies, near-field radar imaging mainly involves performing differential calculations or adaptive filtering based on a background template to obtain radar imaging results for the target scene.

[0004] Because the background templates in the existing technology cannot adapt to complex and ever-changing scenes, the radar signals or image signals extracted by filtering lack precision, resulting in the technical problem of low radar imaging accuracy in the existing technology. Summary of the Invention

[0005] This application provides a near-field MIMO radar data processing method based on multi-level background suppression and target enhancement, in order to improve the technical effect of radar imaging accuracy.

[0006] In a first aspect, embodiments of this application provide a near-field MIMO radar data processing method based on multi-level background suppression and target enhancement, including:

[0007] Obtain the real-time background template and real-time image frames of the target scene;

[0008] Based on the real-time background template and the real-time image frame, the differential image data corresponding to the real-time image frame is calculated;

[0009] Gaussian filtering is applied to the difference image data to obtain Gaussian filtered image data;

[0010] Morphological processing is performed on Gaussian filtered image data to obtain morphological image data;

[0011] Target enhancement processing is performed on morphological image data to obtain target enhanced image data corresponding to real-time image frames;

[0012] Output target augmented image data.

[0013] In one possible implementation, obtaining the real-time background template of the target scene includes:

[0014] Acquire multiple empty field image frames of the target scene in an empty field state;

[0015] An initial background template is generated based on multiple empty field image frames;

[0016] The current image frame of the target scene is acquired at a preset period.

[0017] The initial background template is updated based on the current image frame to obtain the real-time background template.

[0018] In one possible implementation, an initial background template is generated based on multiple empty field image frames, including:

[0019] Generate the imaging results for each empty field image frame;

[0020] The average of multiple imaging results is calculated to obtain the initial background template.

[0021] In one possible implementation, the initial background template is updated based on the current image frame to obtain a real-time background template, including:

[0022] Calculate the normalized difference between the current image frame and the initial background template;

[0023] When the normalization difference is lower than a preset threshold, the current image frame is determined to be an empty field image frame that does not contain the target object;

[0024] The initial background template is updated based on the current image frame to obtain the real-time background template.

[0025] In one possible implementation, differential image data corresponding to the real-time image frame is calculated based on the real-time background template and the real-time image frame, including:

[0026] Calculate the initial residual between the real-time background template and the real-time image frame;

[0027] Based on a preset sliding window length, calculate the average value of multiple frames of image data within the preset sliding window for real-time image frames;

[0028] The difference between the average value and the real-time background template is calculated to obtain the second-order moving average residual.

[0029] The preliminary residual and the second-order moving average residual are weighted and calculated based on preset weighting parameters to obtain the differential image data corresponding to the real-time image frame.

[0030] In one possible implementation, target enhancement processing is performed on the morphological image data to obtain target-enhanced image data corresponding to the real-time image frame, including:

[0031] The morphological image data is sequentially subjected to frequency domain enhancement processing and structural enhancement processing to obtain structurally enhanced image data;

[0032] Calculate the local mean of the neighborhood of each pixel in the structure-enhanced image data;

[0033] Based on each local mean, the consistency adjustment factor for each pixel is calculated;

[0034] Local consistency adjustment is performed on the structure-enhanced image data based on the consistency adjustment factor of each pixel to obtain the target-enhanced image data.

[0035] In one possible implementation, after performing target enhancement processing on the morphological image data to obtain target-enhanced image data corresponding to the real-time image frame, the method further includes:

[0036] The target enhancement image data is normalized to obtain a soft mask;

[0037] The real-time background template is updated based on a soft mask to obtain the updated real-time background template;

[0038] Nonlinear differential fusion weighted calculation is performed based on the initial residual, the second-order moving average residual, and the consistency adjustment factor to obtain nonlinear differential fusion image data;

[0039] Gaussian filtering and morphological processing are performed on the nonlinear differential fusion image data to obtain updated morphological image data.

[0040] Target augmentation processing is performed on morphological image data to obtain target augmented image data corresponding to real-time image frames, including:

[0041] The updated morphological image data is subjected to target augmentation processing to obtain updated target-enhanced image data.

[0042] Secondly, embodiments of this application provide a near-field MIMO radar data processing device based on multi-level background suppression and target enhancement, comprising:

[0043] The acquisition module is used to acquire the real-time background template and real-time image frames of the target scene;

[0044] The first processing module is used to calculate the differential image data corresponding to the real-time image frame based on the real-time background template and the real-time image frame.

[0045] The second processing module is used to perform Gaussian filtering on the differential image data to obtain Gaussian filtered image data.

[0046] The third processing module is used to perform morphological processing on the Gaussian filtered image data to obtain morphological image data.

[0047] The fourth processing module is used to perform target enhancement processing on the morphological image data to obtain target enhanced image data corresponding to the real-time image frame;

[0048] The fifth processing module is used to output target enhancement image data.

[0049] In one possible implementation, the acquisition module is further configured to:

[0050] Acquire multiple empty field image frames of the target scene in an empty field state;

[0051] An initial background template is generated based on multiple empty field image frames;

[0052] The current image frame of the target scene is acquired at a preset period.

[0053] The initial background template is updated based on the current image frame to obtain the real-time background template.

[0054] Optionally, the acquisition module is also used for:

[0055] Generate the imaging results for each empty field image frame;

[0056] The average of multiple imaging results is calculated to obtain the initial background template.

[0057] In one possible implementation, the acquisition module is further configured to:

[0058] Calculate the normalized difference between the current image frame and the initial background template;

[0059] When the normalization difference is lower than a preset threshold, the current image frame is determined to be an empty field image frame that does not contain the target object;

[0060] The initial background template is updated based on the current image frame to obtain the real-time background template.

[0061] In one possible implementation, the first processing module is further configured to:

[0062] Calculate the initial residual between the real-time background template and the real-time image frame;

[0063] Based on a preset sliding window length, calculate the average value of multiple frames of image data within the preset sliding window for real-time image frames;

[0064] The difference between the average value and the real-time background template is calculated to obtain the second-order moving average residual.

[0065] The preliminary residual and the second-order moving average residual are weighted and calculated based on preset weighting parameters to obtain the differential image data corresponding to the real-time image frame.

[0066] In one possible implementation, the fourth processing module is further configured to:

[0067] The morphological image data is sequentially subjected to frequency domain enhancement processing and structural enhancement processing to obtain structurally enhanced image data;

[0068] Calculate the local mean of the neighborhood of each pixel in the structure-enhanced image data;

[0069] Based on each local mean, the consistency adjustment factor for each pixel is calculated;

[0070] Local consistency adjustment is performed on the structure-enhanced image data based on the consistency adjustment factor of each pixel to obtain the target-enhanced image data.

[0071] In one possible implementation, the device further includes a fourth processing module for:

[0072] The target enhancement image data is normalized to obtain a soft mask;

[0073] The real-time background template is updated based on a soft mask to obtain the updated real-time background template;

[0074] Nonlinear differential fusion weighted calculation is performed based on the initial residual, the second-order moving average residual, and the consistency adjustment factor to obtain nonlinear differential fusion image data;

[0075] Gaussian filtering and morphological processing are performed on the nonlinear differential fusion image data to obtain updated morphological image data.

[0076] The updated morphological image data is subjected to target augmentation processing to obtain updated target-enhanced image data.

[0077] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0078] The memory stores the instructions that the computer executes;

[0079] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect above and various possible implementations of the first aspect.

[0080] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and various possible implementations thereof.

[0081] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and various possible implementations thereof.

[0082] This application provides a near-field MIMO radar data processing method based on multi-level background suppression and target enhancement. The method acquires a real-time background template and real-time image frames of the target scene; calculates differential image data corresponding to the real-time image frames based on the real-time background template and real-time image frames, achieving differential processing. Gaussian filtering is applied to the differential image data after differential processing to obtain Gaussian-filtered image data for further clutter suppression. Morphological processing is then applied to the Gaussian-filtered image data to obtain morphological image data, achieving noise removal. Target enhancement processing is performed on the morphological image data to clarify the radar imaging target, obtaining and outputting target-enhanced image data. Compared with existing technologies, this application utilizes differential processing for initial background suppression; Gaussian blurring is used for further clutter suppression; morphological processing removes isolated noise points from the image data; and target enhancement is used to obtain target-enhanced image data, thereby improving the processing accuracy of radar signals. Furthermore, combining the acquired real-time background template with data processing avoids the situation of incompatibility between complex backgrounds and the template, further improving data processing accuracy, thus achieving the technical effect of improving radar imaging accuracy. Attached Figure Description

[0083] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0084] Figure 1 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 1 ;

[0085] Figure 2 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 2 ;

[0086] Figure 3 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 3 ;

[0087] Figure 4 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 4 ;

[0088] Figure 5A schematic diagram of the near-field MIMO radar data processing device based on multi-level background suppression and target enhancement provided in this application;

[0089] Figure 6 A schematic diagram of the structure of the electronic device provided in this application.

[0090] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0091] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0092] First, let's explain the terms involved in the application:

[0093] Multiple-Input Multiple-Output (MIMO): In radar technology, MIMO overcomes the performance limitations of traditional single-antenna or few-antenna radars by coordinating multiple antennas independently transmitting signals at the transmitting end and multiple antennas synchronously receiving signals at the receiving end, thereby achieving higher resolution, anti-jamming capability, and target detection efficiency.

[0094] Field-Programmable Gate Array (FPGA): refers to a programmable logic device whose core feature is that after the chip leaves the factory, users can reconfigure the internal circuit structure of the chip according to specific needs to realize custom digital logic functions.

[0095] In existing technologies, near-field radar imaging involves background suppression and filtering of signals acquired by the radar device using a pre-set background template; then, imaging processing is performed based on the processed signal data to achieve radar imaging.

[0096] However, in radar imaging, the background template update method is not matched with the complexity of the scene being identified; there are cases where the scene changes but the background template does not. At the same time, when filtering or suppressing the obtained radar signal, the data processing accuracy remains at the level of simple clutter suppression, lacking fine data processing. As a result, there is still a lot of interference information in the final imaging result, which leads to the technical problem of low radar imaging accuracy in the existing technology.

[0097] To address the aforementioned technical problems, this application proposes the following technical concept: When acquiring the background template, the latest real-time background template is used; using the acquired real-time background template and real-time image frames, differential calculation is performed to obtain differential image data, achieving preliminary background suppression. Gaussian filtering is then applied to the differential image data for further clutter suppression, resulting in Gaussian-filtered image data. Morphological processing is performed on the Gaussian-filtered image data to remove independent noise points, obtaining morphological image data, thus achieving multi-level suppression and filtering. Target enhancement processing is then performed using the obtained morphological image data to clarify the radar imaging target, thereby obtaining target-enhanced image data representing a clear imaging result. Compared with existing technologies, the use of a real-time background template and multi-level suppression filtering achieves the technical effect of improving radar imaging accuracy.

[0098] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0099] Figure 1 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 1 ,like Figure 1 As shown, the method includes:

[0100] S101. Obtain the real-time background template and real-time image frames of the target scene.

[0101] In this step, the real-time background template refers to the reference benchmark for the background signal or imaging features received by the near-field radar system. A real-time image frame refers to the two-dimensional or three-dimensional image unit in a time series generated by the near-field radar system through signal transmission and reception, converting electromagnetic scattering information in the target scene. Each image frame corresponds to a complete signal acquisition cycle, including information from multiple dimensions such as distance, angle, and velocity in the target scene, and is ultimately presented in a visual form.

[0102] Optionally, the real-time background template can be obtained by retrieving the most recently updated real-time background template from a preset database; alternatively, the real-time background template can be obtained by updating the previously used background template based on the currently acquired targetless image frame.

[0103] For example, a way to acquire real-time image frames can be:

[0104] a1. Design a MIMO array architecture. Specifically: Virtual aperture expansion, using a multiple-transmit, multiple-receive antenna array, generating virtual array elements through time-division multiplexing or frequency-division multiplexing techniques. Spatial calibration, using infrared cameras or electromagnetic simulation to obtain antenna position parameters and correct phase errors caused by array layout.

[0105] a2. Orthogonal waveform generation and transmission. Specifically, orthogonal linear frequency modulated signals are generated for each transmission channel, minimizing cross-correlation peaks through frequency spacing. Trigger signals are generated via FPGA to ensure synchronous transmission of multi-channel signals at the nanosecond level.

[0106] a3. High-speed analog-to-digital converter (ADC) and channel isolation. Specifically, the selection and setting of the high-speed ADC involves using a high-speed ADC with at least twelve bits and setting the full-scale input power to mitigate weak echoes and strong leakage signals. Channel isolation and optimization are achieved through spatial isolation, reverse output of the suppressor, and a digital background cancellation algorithm to improve channel isolation to above a pre-set decibel level.

[0107] a4. Combine the above hardware signal acquisition process to obtain the radar detection signal of the target scene, perform preliminary filtering and correction processing on the radar detection signal, and obtain the real-time image frame of the target scene.

[0108] It should be noted that the specific implementation method for obtaining the real-time background template is as follows. Figure 2 Further explanation will be provided in the embodiments shown, and will not be repeated here.

[0109] S102. Based on the real-time background template and the real-time image frame, calculate the differential image data corresponding to the real-time image frame.

[0110] In this step, differential image data refers to image data with high purity and high signal-to-noise ratio obtained by performing two-level differential processing on real-time image frames based on real-time background templates.

[0111] Alternatively, one possible way to obtain the differential image data is as follows:

[0112] S1021. Calculate the initial residual between the real-time background template and the real-time image frame.

[0113] In this step, the preliminary residual refers to the preliminary residual matrix obtained by performing a first-level difference between the current real-time image frame and the real-time background template.

[0114] For example, the calculation method for the first-order difference is shown in Formula 1:

[0115]

[0116] in, This refers to the initial difference result of the real-time image frame, i.e., the calculated preliminary residual; this difference can effectively eliminate stable strong scattering components of the background from the data signal of the real-time image frame. k refers to the index or sequence number of the real-time image frame, and x and y refer to the spatial coordinate dimensions in the real-time image frame, corresponding to the two-dimensional pixel coordinates of near-field radar imaging, which can be understood as the coordinates of the pixels in the real-time image frame. This refers to the data matrix corresponding to real-time image frames. This refers to the data matrix corresponding to the real-time background template.

[0117] S1022. Based on the preset sliding window length, calculate the average value of multiple frames of image data within the preset sliding window for real-time image frames.

[0118] In this step, in order to suppress occasional noise and strong interference pulses, it is necessary to introduce frame-level sliding smoothing with a preset sliding window length, and calculate the average image echo of the preset sliding window length, that is, the average value between multiple frames of image data within the preset sliding window.

[0119] For example, the average value of image data is calculated as shown in Formula 2:

[0120]

[0121] in, This refers to the time-averaged echo of the most recent L frames, which is the average value among multiple frames of image data; L refers to the length of the preset sliding window; k refers to the sequence number or index of the real-time image frame. This refers to the data matrix corresponding to the i-th image frame.

[0122] S1023. Perform differential calculation on the average value and the real-time background template to obtain the second-level moving average residual.

[0123] In this step, the purpose of performing differential calculation between the calculated average value and the real-time background template is to further offset dynamic background residue and improve the signal-to-noise ratio of the imaging target.

[0124] For example, the second-order moving average residual is calculated as shown in Formula 3:

[0125]

[0126] in, This refers to the second-order difference calculation result of the real-time image frame, i.e., the second-order moving average residual. The explanations of the other parameters are shown in Formulas 1 and 2 above.

[0127] S1024. Based on preset weighting parameters, perform weighted calculations on the initial residual and the second-level moving average residual to obtain the differential image data corresponding to the real-time image frame.

[0128] In this step, the purpose of weighted calculation is to balance background suppression and target integrity. The initial residual calculation process removes most of the background information, but there are residual dynamic interferences. The calculation of the second-order moving average residual further eliminates the residual dynamic background, but there may be over-smoothing of the target edges. Therefore, it is necessary to perform weighted adjustment on the two residual results to obtain the final difference image data.

[0129] Optionally, the design of the preset weighting parameters in this step needs to match the radar imaging purpose of the target scene; when the target scene is a dynamic near-field radar imaging scene, the weight of the second-order moving average residual needs to be increased to prioritize dynamic interference; when the target scene is a static scene, the weight of the initial residual can be increased to preserve the fine structure of the radar imaging target.

[0130] For example, the differential image data is calculated in the manner shown in Equation 4:

[0131]

[0132] in, This refers to the differential image data corresponding to the real-time image frame; This refers to the initial residual corresponding to the image frame; This refers to the second-order moving average residual corresponding to the real-time image frame; w refers to the preset weighting parameter; and k refers to the sequence number or index of the real-time image frame.

[0133] S103. Perform Gaussian filtering on the differential image data to obtain Gaussian filtered image data.

[0134] In this step, the Gaussian filtering process is as follows: Gaussian kernels are constructed with different standard deviations, and convolution is performed on the difference image data after background suppression. The convolutional data corresponding to different Gaussian kernels are then subtracted to retain the high-pass component, resulting in Gaussian-filtered image data. Specifically, Gaussian filtering refers to convolution using Gaussian kernels and difference calculation using subtraction to obtain the final Gaussian difference result.

[0135] For example, two standard deviations are designed as follows: A Gaussian kernel is constructed based on these two standard deviations, and the difference image data is convolved using the convolution method shown in Equation 5:

[0136]

[0137] in, and This refers to the parameter being Two two-dimensional Gaussian kernels; and This refers to smoothed image data after being convolved with two Gaussian kernels; This refers to differential image data, where (x,y) refers to the pixel coordinates of the differential image data.

[0138] The two-dimensional Gaussian kernel is constructed as shown in Formula 6:

[0139]

[0140] in, This refers to the Gaussian kernel parameter, i.e., the standard deviation; u and v refer to two-dimensional spatial coordinates, representing the relative position of the Gaussian kernel within the local processing window. This refers to the weight value of the two-dimensional Gaussian kernel at coordinates (u,v); This refers to the normalization coefficient, which is used to ensure that the sum of the weights of the Gaussian kernel is 1. This refers to the natural exponential function.

[0141] For example, the method of subtracting the data after convolution corresponding to different Gaussian kernels and retaining the high-pass component to obtain Gaussian filtered image data is shown in Equation 7:

[0142]

[0143] in, This refers to the final calculated Gaussian filtered image data; and This refers to smoothed image data after being convolved with two Gaussian kernels.

[0144] It should be noted that the Gaussian filtered image data obtained in this way can remove soft backgrounds that may still exist with slow spatial changes, such as wall reflections, floor reflections, or instrument clutter.

[0145] S104. Perform morphological processing on the Gaussian filtered image data to obtain morphological image data.

[0146] In this step, the morphological processing involves selecting a pre-defined structural unit in both the x and y dimensions, performing an erosion operation on the Gaussian-filtered image data, and then performing a dilation operation on the erosion result to obtain the morphologically processed result. Morphological processing is a nonlinear filtering technique that adjusts the spatial shape and structure of an image based on structural units. The erosion operation in morphological processing refers to the contraction operation, used to extract foreground regions in the image data that cannot be fully contained within the structural unit through scanning. The dilation operation, the opposite of erosion, mainly uses the scanning of structural units to convert background regions that overlap with the structural unit into foreground regions.

[0147] For example, a 3×3 structural unit is used. Morphological processing is performed using the etching operation method shown in Formula 8:

[0148]

[0149] Here, E refers to the structural unit. This refers to the result of erosion operation on Gaussian filtered image data. (x,y) refers to the absolute spatial coordinates of the eroded image, and (i,j) belongs to structural unit E and refers to the relative coordinates within the structural unit. This refers to the final calculated Gaussian filtered image data; k refers to the index or sequence number of the real-time image data.

[0150] The expansion operation is performed as shown in Formula 9:

[0151]

[0152] in, This refers to the result of morphological processing, i.e., morphological image data; the explanation of the other parameters is given in Formula 8 above.

[0153] It should be noted that after morphological processing, isolated background noise in the image data is removed, while the target subject in radar imaging remains unaffected, further improving recognition accuracy.

[0154] S105. Perform target enhancement processing on the morphological image data to obtain target enhanced image data corresponding to the real-time image frame.

[0155] Alternatively, one possible way to obtain target-enhanced image data is as follows:

[0156] S1051. Perform frequency domain enhancement processing and structural enhancement processing on the morphological image data in sequence to obtain structurally enhanced image data.

[0157] Alternatively, one possible way to obtain structure-enhanced image data is as follows:

[0158] b1. For each pixel in the morphological image data, extract the neighborhood corresponding to that pixel.

[0159] For example, for morphological image data The neighborhood corresponding to the pixel is extracted as shown in Formula 10:

[0160]

[0161] in, This refers to the data extracted from the neighborhood of coordinates (x,y) in morphological image data; (i,j) refers to the coordinates of the pixels in the neighborhood.

[0162] In this step, the neighborhood extraction can be performed by: setting the neighborhood shape, such as a square neighborhood, a circular neighborhood, or a cross neighborhood; setting the neighborhood size; extracting other pixels in the neighborhood centered on the pixel according to the set neighborhood shape and neighborhood size; and using the combination of extracted pixels as the neighborhood of the central pixel.

[0163] b2. Perform Fourier transform on each pixel in the neighborhood to obtain the local power spectrum corresponding to each neighborhood.

[0164] In this step, the Fourier transform refers to performing a Fourier transform after applying a two-dimensional weighted window to the neighborhood. The method of two-dimensional Fourier transform is shown in Equation 11:

[0165]

[0166] in, This refers to the local power spectrum within the domain; This refers to the data within the neighborhood extracted from morphological image data with coordinates (x, y); This refers to the two-dimensional Fourier transform function.

[0167] b3. For each local power spectrum, calculate the spectral centroid coordinates and radial broadening of each local power spectrum.

[0168] In this step, the spectral centroid is the power-weighted frequency center in the local power spectrum; radial broadening is the weighted average of the radial distances of each frequency component in the local power spectrum to the spectral centroid.

[0169] For example, the spectral centroid coordinates and radial broadening of each local power spectrum are calculated in the manner shown in Equation 12:

[0170]

[0171] Formula 12 is used to calculate the radial broadening of the neighborhood centered at the spatial domain coordinates (x, y). (x, y) corresponds to the center coordinates of the neighborhood in the morphologically processed image. This refers to mapping the frequency domain abscissa u of the local power spectrum to morphological image data. The x-coordinate in the diagram; This refers to mapping the frequency domain abscissa v of the local power spectrum to morphological image data. The vertical coordinate in the diagram. This refers to the lateral coordinates of the centroid of the local power spectrum in the morphological image; This refers to the vertical coordinate of the centroid of the local power spectrum in the morphological image. This refers to the power value of the local power spectrum at the frequency domain coordinates (u,v). This refers to radial widening; This refers to the centroid coordinates.

[0172] b4. Normalize the radial expansion of multiple neighborhoods to obtain the weights corresponding to each neighborhood.

[0173] In this step, normalization refers to mapping the calculated original radial broadening to a pre-defined range to avoid over- or under-enhancement due to differences in the magnitude of the original radial broadening values.

[0174] b5. Based on the weights corresponding to each neighborhood, the morphological image data is weighted to obtain frequency domain enhanced image data.

[0175] In this step, the weighted calculation method is as shown in Formula 13:

[0176]

[0177] in, This refers to frequency-domain enhanced image data obtained based on spectrum broadening; This refers to morphological image data. This refers to the weight corresponding to each neighborhood with center coordinates (x, y).

[0178] b6. Calculate the local second-order difference of each pixel based on frequency domain enhanced image data.

[0179] In this step, the purpose of calculating the local second-order difference for each pixel is to calculate the local second-order difference along the two-dimensional spatial dimension based on the frequency domain enhanced image data, in order to capture the amplitude of the transition in the edge region.

[0180] For example, the local second-order difference is calculated as shown in Equation 14:

[0181]

[0182] in, , These are the local second-order difference results for the horizontal and vertical directions, respectively. This refers to frequency-domain enhanced image data obtained based on spectrum broadening; and This refers to the one-dimensional difference component, which represents the pixel interval during difference calculation. (x+△x,y) and (x-△x,y) refer to the right and left adjacent pixels of the target pixel (x,y); similarly, (x,y+△y) and (x,y-△y) refer to the bottom and top adjacent pixels of the target pixel (x,y).

[0183] It should be noted that the purpose of calculating the second-order difference is to provide edge transition data in different directions for the subsequent edge activation process, so as to avoid missing the edge of the radar imaging target by the single-direction difference.

[0184] b7. Based on the local second-order difference of each pixel, edge activation is calculated to obtain the edge activation value corresponding to each pixel.

[0185] In this step, the edge activation values ​​are calculated using the spatial edge activation function, which integrates the second-order differences between two directions into edge strengths in all directions, ensuring that no edge in any direction is missed.

[0186] For example, the edge activation function is shown in Equation 15:

[0187]

[0188] Where E(x,y) refers to the output value of the spatial edge activation function, representing the edge intensity in all directions at pixel (x,y), i.e., the edge activation value; and These refer to the squares of the second-order differences in the x and y directions, respectively.

[0189] It should be noted that the process of calculating the edge activation value is used to convert the edge transitions in different directions into edge intensity in a single dimension, which facilitates subsequent unified judgment of whether it is the edge of a radar imaging target.

[0190] b8. Based on the edge activation value of each pixel and preset control parameters, the structure-enhanced image data is calculated.

[0191] In this step, the process of obtaining structure-enhanced image data is as follows: constructing a local response amplification signal based on the spatial edge activation function, and calculating structure-enhanced image data with significant enhancement of spatial edges using preset control parameters.

[0192] For example, spatial edge enhancement can be achieved as shown in Equation 16:

[0193]

[0194] in, This refers to the final output structure-enhanced image data; This refers to frequency domain enhanced image data; This refers to the parameter that controls the magnitude of edge enhancement. This refers to the parameter that controls the sensitivity of the enhancement term to changes in edge intensity; tanh(·) refers to the hyperbolic tangent function. E(x,y) refers to the edge intensity.

[0195] S1052. Calculate the local mean of the neighborhood of each pixel in the structure-enhanced image data.

[0196] In this step, to prevent the signal of the structure-enhanced image from spreading into the background, it is necessary to calculate the local mean of each enhanced pixel in its neighborhood to reflect the overall gray level or radar echo level of the neighborhood around that pixel.

[0197] S1053. Based on each local mean, calculate the consistency adjustment factor for each pixel.

[0198] In this step, the consistency adjustment factor is calculated by comparing the enhanced pixel value with the mean value of the local neighborhood to determine whether the pixel belongs to the real target area, thereby dynamically adjusting the enhancement intensity.

[0199] For example, the consistency adjustment factor is calculated as shown in Formula 17:

[0200]

[0201] in, This refers to the consistency adjustment factor corresponding to the pixel (x,y), which is a parameter between 0 and 1; This refers to structure-enhanced image data; This refers to the local mean of the structure-enhanced image data within a range of u at a scale of K; where K refers to the scale of the neighborhood. This refers to adding a very small positive number to the enhanced pixel value. , as a denominator, is used to avoid abnormal division by zero or extremely small numbers.

[0202] S1054. Local consistency adjustment is performed on the structure enhancement image data based on the consistency adjustment factor of each pixel to obtain the target enhancement image data.

[0203] In this step, local consistency adjustment refers to using a consistency adjustment factor to perform target region enhancement preservation and background region enhancement suppression on the enhanced structure-enhanced image data, outputting the final target-enhanced image data. The target region refers to the area where the radar-imported target is located.

[0204] For example, the consistency adjustment method is shown in Equation 18:

[0205]

[0206] in, This refers to target augmentation image data. This refers to structure-enhanced image data. This refers to the consistency adjustment factor. x and y refer to the coordinate information of the pixel.

[0207] S106, Output target enhanced image data.

[0208] In this step, the target augmented image data can be output in two ways: visually displaying the target augmented image data on a monitor; and simultaneously saving the target augmented image data in a pre-set database. When the target augmented image data needs to participate in further data processing and calculation, it can be synchronously sent to the corresponding data processing platform. Target image augmentation data refers to the target recognition results of near-field radar imaging. The output method of this data is not entirely consistent in different recognition scenarios, and it is necessary to select an appropriate output and display method based on different recognition scenarios.

[0209] This application provides a near-field MIMO radar data processing method based on multi-level background suppression and target enhancement. This method acquires a real-time background template and real-time image frames of the target scene; calculates differential image data corresponding to the real-time image frames based on the real-time background template and real-time image frames, achieving differential processing. Gaussian filtering is applied to the differential image data after differential processing to obtain Gaussian-filtered image data for further clutter suppression. Morphological processing is then applied to the Gaussian-filtered image data to obtain morphological image data, achieving noise removal. Target enhancement processing is performed on the morphological image data to clarify the radar imaging target, obtaining and outputting target-enhanced image data. Compared with existing technologies, this application utilizes differential processing for initial background suppression; Gaussian blurring is used for further clutter suppression; morphological processing removes isolated noise points from the image data; and target enhancement is used to obtain target-enhanced image data, thereby improving the processing accuracy of radar signals. Furthermore, combining the acquired real-time background template with data processing avoids mismatches between complex backgrounds and the template, further improving data processing accuracy, thus achieving the technical effect of improving radar imaging accuracy.

[0210] Figure 2 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 2 This embodiment refers to the above. Figure 1 The acquisition of the real-time background template in step S101 of the illustrated embodiment will be further explained, as follows: Figure 2 As shown, the method includes:

[0211] S201. Acquire multiple empty field image frames of the target scene in an empty field state.

[0212] In this step, the empty field state refers to a clean environment in the target scene where there is no target object to be detected, and only background elements are present.

[0213] For example, an empty field state can be: no personnel in the security checkpoint, no workpiece defects in the industrial inspection area, and no vehicles or pedestrians in an autonomous driving scenario.

[0214] An airfield image frame refers to a series of original images acquired by a near-field MIMO radar in an airfield state according to the normal imaging process.

[0215] In this step, the purpose of acquiring empty field image frames is to collect pure background data without target interference, so as to avoid the introduction of target signals into the initial background template; at the same time, by acquiring multiple frames, the background data is accumulated to provide samples for subsequent filtering and suppression processing.

[0216] S202. Generate an initial background template based on multiple empty field image frames.

[0217] In this step, the initial background template serves as a stable reference for the background signal under the spatiotemporal field state.

[0218] Alternatively, one possible implementation for generating the initial background template is as follows:

[0219] S2021. Generate the imaging results corresponding to each empty field image frame.

[0220] In this step, the imaging result refers to the clean background data obtained after radar signal preprocessing of a single frame of original airfield image, rather than the original echo signal.

[0221] For example, the imaging result of the airfield image frame can be generated as follows: amplitude calibration and phase correction are performed on the raw echo signal acquired by the MIMO radar; a range-angle two-dimensional grayscale image of the airfield image frame is generated; the two-dimensional grayscale image is filtered using Gaussian filtering to suppress high-frequency random noise and retain low-frequency stable components in the background. The processed signal is then converted into a uniform grayscale format to obtain the single-image airfield imaging result.

[0222] S2022. Calculate the mean of multiple imaging results to obtain the initial background template.

[0223] In this step, the initial background template refers to a two-dimensional matrix obtained by calculating the pixel-level mean of the imaging results of all empty field image frames. It represents the average value of the background signal in the empty field state and serves as the initial reference for subsequent real-time template updates.

[0224] For example, the initial background template is calculated as shown in Formula 19:

[0225]

[0226] in, This refers to the initial background template; N refers to the number of empty field image frames; n refers to the sequence number of the empty field image frames. This refers to the imaging result of the nth empty field image frame; (x,y) refers to the coordinates of the pixel in the imaging result.

[0227] S203. Acquire the current image frame of the target scene within a preset period.

[0228] In this step, the preset period refers to the time interval at which the radar system periodically acquires scene images, which needs to be matched with the dynamic nature of the scene. The current image frame refers to the imaging result of the target scene in real time acquired by the radar within the preset period, which may include the identified target or only the background.

[0229] The purpose of this step is to dynamically monitor the real-time changes of the target scene and provide the latest update data for real-time updates of the background template; at the same time, it ensures the timeliness of the background template and adapts to complex and ever-changing target scenes.

[0230] For example, in a static scenario such as industrial inspection, the preset cycle length can be set to 500ms or more; in a dynamic scenario such as safety and autonomous driving, the preset cycle length can be set to between 10ms and 100ms.

[0231] S204. Update the initial background template according to the current image frame to obtain the real-time background template.

[0232] Alternatively, one possible way to obtain a real-time background template is as follows:

[0233] S2041. Calculate the normalized difference between the current image frame and the initial background template.

[0234] In this step, the normalized difference refers to the difference quantization value obtained by normalizing the grayscale value of the pixel-level absolute difference between the current image frame and the initial template. This is used to avoid unfair comparisons caused by the difference in absolute grayscale values ​​of different pixels, where the difference in high grayscale pixels is large and the difference in low grayscale pixels is small.

[0235] S2042. When the normalized difference is lower than the preset threshold, the current image frame is determined to be an empty field image frame that does not contain the target object.

[0236] In this step, the preset threshold is used to distinguish between two cases: the current image frame is an empty image frame and the current image frame contains the detected target. An empty image frame that does not contain the target refers to a normalized difference value lower than the preset threshold, meaning that the background and the initial template are at the same height and there is no target signal interference.

[0237] For example, the normalized difference calculation and judgment method is shown in Formula 20:

[0238]

[0239] in, This refers to the grayscale value of the current image frame at (x, y). This refers to the grayscale value of the initial background template at (x, y); This refers to the preset threshold; k refers to the sequence number of the current image frame.

[0240] S2043. Update the initial background template based on the current image frame to obtain the real-time background template.

[0241] In this step, the real-time background template refers to the dynamic background reference obtained by iteratively updating the initial background template through targetless empty field image frames.

[0242] For example, the real-time background template can be obtained by using an exponentially weighted average calculation method, as shown in Formula 21:

[0243]

[0244] in, This refers to the real-time background template updated based on the current image frame; This refers to the last updated background template; the last updated background template can be the initial background template obtained from the first calculation, or the background template after the update of the image frame corresponding to the k-1 sequence number; it represents the initial background template during the first update. This refers to updating the weight coefficient, and the value range can be set to 0.8 to 0.99. This refers to the current image frame.

[0245] Figure 3 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 3 Based on the above embodiments, this application also provides a method for feedback optimization based on target enhancement image data, such as... Figure 3 As shown, the method includes:

[0246] S301. Normalize the target enhancement image data to obtain a soft mask.

[0247] In this step, the target enhanced image data refers to the data after enhancement processing, characterized by a significantly higher grayscale value in the target area compared to the background area. The soft mask refers to a pixel-level weight matrix generated through normalization, with values ​​ranging from [0,1]. It is used to quantify the probability that a pixel belongs to the target area; the soft mask for the target area is close to 1, and the soft mask for the background area is close to 0.

[0248] For example, the normalization process is used to obtain the soft mask as shown in Equation 22:

[0249]

[0250] in, This refers to a soft mask; This refers to the target augmented image data, where (x, y) refers to the coordinates of a pixel in the target augmented image data. Formula 22 calculates the soft mask value for each pixel by dividing the gray value of each pixel by the global maximum gray value.

[0251] S302. Update the real-time background template based on the soft mask to obtain the updated real-time background template.

[0252] In this step, updating the real-time background template refers to updating the real-time background template from the previous iteration or the previous moment.

[0253] For example, the real-time background template is updated as shown in Formula 23:

[0254]

[0255] in, This refers to the updated live background template. This refers to the last updated live background template; This refers to the stability weight, which can be set to [0.7, 0.95] and is used to control the degree of influence of the last updated template. This refers to the current image frame; that is, the original scene image acquired in the current period. This refers to the soft mask corresponding to the target augmentation image data.

[0256] It should be noted that the parts of the signal with stronger echoes after enhancement are more likely to contain the target. The echo signal mainly comes from the target rather than background clutter and noise signals. Therefore, background suppression should be less applied to this part of the data, hence the soft mask update of the background template.

[0257] S303. Based on the initial residual, the second-order moving average residual, and the consistency adjustment factor, nonlinear differential fusion weighted calculation is performed to obtain nonlinear differential fusion image data.

[0258] In this step, the nonlinear differential fusion image data obtained refers to target-driven consistency-adjusted differential fusion. In the above... Figure 1 In the weighted calculation of the preliminary residual and the second-order moving average residual in step S102 of the embodiment shown, fixed weights are used. In scenarios such as security checks, the frequency of contraband detection targets is low. Therefore, the calculation of the second-order moving average residual will weaken the strength of the target signal to a certain extent. In this case, the updated first-order differential calculation should be used in the echo signal where the target exists.

[0259] To design adaptive weights, the adjustment function needs to be constructed using the consistency adjustment factor in the target enhancement process, as shown in Equation 24:

[0260]

[0261] in, This refers to the fusion weight; w(x,y) refers to the consistency adjustment factor. This refers to a tiny positive number, used to avoid the problem of not being able to calculate when the consistency adjustment factor is 0.

[0262] Further nonlinear fusion is performed based on the generated fusion weights, and the fusion method is shown in Equation 25:

[0263]

[0264] in, This refers to the preliminary residual calculated in step S102 above; This refers to the second-order moving average residual calculated in step S102 above; This refers to the fusion weight; This refers to nonlinear differential fusion image data.

[0265] S304. Gaussian filtering and morphological processing are performed on the nonlinear differential fusion image data to obtain updated morphological image data.

[0266] In this step, Gaussian filtering refers to convolving the nonlinear difference fusion image data using a two-dimensional Gaussian kernel and smoothing high-frequency noise. Morphological processing refers to image operations based on mathematical morphology, including erosion followed by dilation, to obtain updated morphological image data.

[0267] It should be noted that the Gaussian filtering and morphological processing methods used in this step can be referenced from [the relevant documentation / reference]. Figure 1 The illustrated embodiment.

[0268] S305. Perform target enhancement processing on the updated morphological image data to obtain updated target enhanced image data.

[0269] In this step, target augmentation processing is performed by reusing... Figure 1 The enhancement strategy of the illustrated embodiment further amplifies the grayscale difference between the target and the background in the morphological image data. The updated target-enhanced image data refers to the image data that has undergone secondary enhancement, where the grayscale value of the target region is significantly higher than that of the background region, providing high-contrast input for the final target detection.

[0270] Figure 4 A flowchart illustrating the near-field MIMO radar data processing method based on multi-level background suppression and target enhancement provided in this application. Figure 4 ,like Figure 4 As shown, the method includes:

[0271] A1. Background template generation and updating.

[0272] Specifically, before performing MIMO radar imaging, a period of time of airfield image frames is acquired, and an initial background template is generated based on these airfield image frames. Real-time airfield image frames are then acquired, and the initial background template is updated using these real-time airfield image frames to obtain the real-time background template.

[0273] A2. Perform two-level differential fusion on the real-time image frame and the real-time background template to obtain differential image data.

[0274] Specifically, the preliminary residual and the second-order moving average residual are calculated, and the differential image data is obtained by weighted fusion.

[0275] A3. Based on the difference image data, Gaussian filtering and morphological processing are performed to remove noise, resulting in morphological image data.

[0276] A4. Target enhancement processing is performed based on the spectral broadening features of morphological image data to obtain frequency domain enhanced image data.

[0277] A5. Structural enhancement is performed on the spatial edges of the frequency domain enhanced image data to obtain structurally enhanced image data.

[0278] A6. Local consistency processing is performed on the structure-enhanced image data to obtain the target-enhanced image data.

[0279] A7. Output target enhanced image data.

[0280] A8. Drive real-time background template update based on target-enhanced image data, and execute step A1.

[0281] A9. Based on target-enhanced image data-driven adaptive differential fusion, execute step A2.

[0282] Figure 5 A schematic diagram of the near-field MIMO radar data processing device based on multi-level background suppression and target enhancement provided in this application is shown below. Figure 5 As shown, the near-field MIMO radar data processing device based on multi-level background suppression and target enhancement provided in this embodiment includes:

[0283] The acquisition module 501 is used to acquire the real-time background template and real-time image frames of the target scene;

[0284] The first processing module 502 is used to calculate the differential image data corresponding to the real-time image frame based on the real-time background template and the real-time image frame.

[0285] The second processing module 503 is used to perform Gaussian filtering on the differential image data to obtain Gaussian filtered image data.

[0286] The third processing module 504 is used to perform morphological processing on the Gaussian filtered image data to obtain morphological image data.

[0287] The fourth processing module 505 is used to perform target enhancement processing on the morphological image data to obtain target enhancement image data corresponding to the real-time image frame;

[0288] The fifth processing module 506 is used to output target enhancement image data.

[0289] Alternatively, in one possible implementation, the acquisition module 501 is further configured to:

[0290] Acquire multiple empty field image frames of the target scene in an empty field state;

[0291] An initial background template is generated based on multiple empty field image frames;

[0292] The current image frame of the target scene is acquired at a preset period.

[0293] The initial background template is updated based on the current image frame to obtain the real-time background template.

[0294] Optionally, the acquisition module 501 is also used for:

[0295] Generate the imaging results for each empty field image frame;

[0296] The average of multiple imaging results is calculated to obtain the initial background template.

[0297] Alternatively, in one possible implementation, the acquisition module 501 is further configured to:

[0298] Calculate the normalized difference between the current image frame and the initial background template;

[0299] When the normalization difference is lower than a preset threshold, the current image frame is determined to be an empty field image frame that does not contain the target object;

[0300] The initial background template is updated based on the current image frame to obtain the real-time background template.

[0301] Optionally, in one possible implementation, the first processing module 502 is further configured to:

[0302] Calculate the initial residual between the real-time background template and the real-time image frame;

[0303] Based on a preset sliding window length, calculate the average value of multiple frames of image data within the preset sliding window for real-time image frames;

[0304] The difference between the average value and the real-time background template is calculated to obtain the second-order moving average residual.

[0305] The preliminary residual and the second-order moving average residual are weighted and calculated based on preset weighting parameters to obtain the differential image data corresponding to the real-time image frame.

[0306] Optionally, in one possible implementation, the fourth processing module 505 is further configured to:

[0307] The morphological image data is sequentially subjected to frequency domain enhancement processing and structural enhancement processing to obtain structurally enhanced image data;

[0308] Calculate the local mean of the neighborhood of each pixel in the structure-enhanced image data;

[0309] Based on each local mean, the consistency adjustment factor for each pixel is calculated;

[0310] Local consistency adjustment is performed on the structure-enhanced image data based on the consistency adjustment factor of each pixel to obtain the target-enhanced image data.

[0311] Optionally, in one possible implementation, the device further includes a fourth processing module 505, for:

[0312] The target enhancement image data is normalized to obtain a soft mask;

[0313] The real-time background template is updated based on a soft mask to obtain the updated real-time background template;

[0314] Nonlinear differential fusion weighted calculation is performed based on the initial residual, the second-order moving average residual, and the consistency adjustment factor to obtain nonlinear differential fusion image data;

[0315] Gaussian filtering and morphological processing are performed on the nonlinear differential fusion image data to obtain updated morphological image data.

[0316] The updated morphological image data is subjected to target augmentation processing to obtain updated target-enhanced image data.

[0317] The apparatus provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0318] Figure 6 A schematic diagram of the structure of the electronic device provided in this application. Figure 6 As shown, the electronic device provided in this embodiment includes at least one processor 601 and a memory 602. Optionally, the device further includes a communication component 603. The processor 601, memory 602, and communication component 603 are connected via a bus 604.

[0319] In the specific implementation process, at least one processor 601 executes computer execution instructions stored in memory 602, causing at least one processor 601 to execute the above-mentioned near-field MIMO radar data processing method or approach based on multi-level background suppression and target enhancement.

[0320] The specific implementation process of processor 601 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0321] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0322] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0323] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0324] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0325] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0326] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0327] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0328] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components 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 indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0329] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0330] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0331] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0332] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0333] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A near-field MIMO radar data processing method based on multi-level background suppression and target enhancement, characterized in that, include: Obtain the real-time background template and real-time image frames of the target scene; Based on the real-time background template and the real-time image frame, differential image data corresponding to the real-time image frame is calculated; The differential image data is subjected to Gaussian filtering to obtain Gaussian filtered image data; Morphological processing is performed on the Gaussian filtered image data to obtain morphological image data; The morphological image data is subjected to target enhancement processing to obtain target enhanced image data corresponding to the real-time image frame; Output the enhanced image data of the target.

2. The method according to claim 1, characterized in that, The process of obtaining the real-time background template for the target scene includes: Acquire multiple empty field image frames of the target scene in an empty field state; An initial background template is generated based on the multiple empty field image frames; The current image frame of the target scene is acquired at a preset period; The initial background template is updated based on the current image frame to obtain the real-time background template.

3. The method according to claim 2, characterized in that, The process of generating an initial background template based on the plurality of empty field image frames includes: Each of the empty field image frames is then used to generate an imaging result. The average of multiple imaging results is calculated to obtain the initial background template.

4. The method according to claim 2, characterized in that, The step of updating the initial background template according to the current image frame to obtain the real-time background template includes: Calculate the normalized difference between the current image frame and the initial background template; When the normalized difference is lower than a preset threshold, the current image frame is determined to be an empty field image frame that does not contain the target object; The initial background template is updated based on the current image frame to obtain the real-time background template.

5. The method according to claim 4, characterized in that, The step of calculating the differential image data corresponding to the real-time image frame based on the real-time background template and the real-time image frame includes: Calculate the preliminary residual between the real-time background template and the real-time image frame; Based on a preset sliding window length, calculate the average value of multiple image data frames within the preset sliding window for the real-time image frame; The difference between the average value and the real-time background template is calculated to obtain the second-order moving average residual; The preliminary residual and the second-order moving average residual are weighted and calculated based on preset weighting parameters to obtain the differential image data corresponding to the real-time image frame.

6. The method according to claim 5, characterized in that, The step of performing target enhancement processing on the morphological image data to obtain target enhanced image data corresponding to the real-time image frame includes: The morphological image data is subjected to frequency domain enhancement processing and structural enhancement processing in sequence to obtain structurally enhanced image data; Calculate the local mean of the neighborhood to which each pixel belongs in the structure-enhanced image data; Based on each local mean, the consistency adjustment factor for each pixel is calculated; The structure-enhanced image data is locally consistent by adjusting the consistency factor of each pixel to obtain the target-enhanced image data.

7. The method according to claim 6, characterized in that, After performing target enhancement processing on the morphological image data to obtain target enhanced image data corresponding to the real-time image frame, the method further includes: The target enhanced image data is normalized to obtain a soft mask; The real-time background template is updated based on the soft mask to obtain the updated real-time background template; Based on the initial residual, the second-order moving average residual, and the consistency adjustment factor, nonlinear differential fusion weighted calculation is performed to obtain nonlinear differential fusion image data; The nonlinear differential fusion image data is subjected to Gaussian filtering and morphological processing respectively to obtain the updated morphological image data; The step of performing target enhancement processing on the morphological image data to obtain target enhanced image data corresponding to the real-time image frame includes: The updated morphological image data is subjected to target enhancement processing to obtain the updated target-enhanced image data.

8. A near-field MIMO radar data processing device based on multi-level background suppression and target enhancement, characterized in that, include: The acquisition module is used to acquire the real-time background template and real-time image frames of the target scene; The first processing module is used to calculate differential image data corresponding to the real-time image frame based on the real-time background template and the real-time image frame. The second processing module is used to perform Gaussian filtering on the differential image data to obtain Gaussian filtered image data. The third processing module is used to perform morphological processing on the Gaussian filtered image data to obtain morphological image data; The fourth processing module is used to perform target enhancement processing on the morphological image data to obtain target enhanced image data corresponding to the real-time image frame; The fifth processing module is used to output the target enhanced image data.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images