Sparse array time domain convolution combined backward projection imaging method and system

Abstract

The invention provides a sparse array time domain convolution combined backward projection imaging method and system, relates to the field of microwave millimeter wave imaging, and solves the technical problem of low imaging precision caused by inaccurate system modeling, multidimensional signal mismatch and strong clutter interference in a microwave millimeter wave three-dimensional imaging system. The method comprises the following steps: a microwave millimeter wave sparse array imaging system obtains a back scattering echo signal from a target, background cancellation operation is carried out on the back scattering echo signal and a background echo signal to obtain a background cancellation echo signal, and meanwhile, an array-frequency dimension two-dimensional time domain convolution filter is generated; after time domain convolution operation is carried out on the background cancellation echo signal in the motion scanning dimension and the frequency dimension, integral calculation is carried out in the frequency dimension to obtain a motion-distance two-dimensional focusing signal, imaging grids are divided in the array dimension and the distance dimension, and the array-distance two-dimensional focusing signal is reconstructed by adopting a backward projection reconstruction method; the method is used in the microwave and millimeter wave imaging process.

Description

Technical Field

[0001] This application relates to the field of microwave and millimeter-wave imaging, and in particular to a sparse array temporal convolution joint back projection imaging method and system. Background Technology

[0002] Microwave, millimeter-wave, and terahertz imaging technologies have been widely applied in aerospace remote sensing, meteorological monitoring, automotive radar, through-wall detection, human body security inspection, and industrial non-destructive testing due to their advantages such as non-ionizing, high penetration, and high safety. Compared with ionizing radiation technologies such as X-rays, microwave and millimeter-wave imaging is safer for operators and has promising engineering application prospects. However, actual production line environments are often limited by installation space and cost, requiring the use of sparse, irregular arrays, such as arc, polygonal, or linear configurations. Traditional time-domain correlation or standard back projection algorithms suffer from low computational efficiency, difficulty in parallelization, and slow imaging speed when processing such arrays, making it difficult to meet the needs of real-time online inspection. Summary of the Invention

[0003] This application provides a sparse array temporal convolution joint back projection imaging method and system, which solves the technical problem of low imaging accuracy caused by inaccurate system modeling, multidimensional signal mismatch and strong clutter interference in microwave and millimeter wave three-dimensional imaging systems.

[0004] To achieve the above objectives, this application adopts the following technical solution: Firstly, a sparse array temporal convolutional joint backprojection imaging method is provided, including: The backscattered echo signal of the object to be detected is acquired and then background cancellation processing is performed with the pre-acquired background echo signal to obtain the background cancellation echo signal. The maximum detection range in the distance dimension is determined based on the bandwidth and number of frequency points of the imaging system, and a distance grid is obtained based on the maximum detection range. A motion-frequency dimension spatial matching filter is obtained based on the distance and motion scan dimension coordinates corresponding to each distance grid point. The imaging system is a microwave millimeter-wave imaging system. Based on the background cancellation echo signal and the motion-frequency dimension spatial matching filter, the motion scanning-distance dimension focusing signal is obtained; The imaging grid is obtained in the array dimension and distance dimension. The transmit slant range and receive slant range from each array element position to each imaging grid point are analyzed, and the array dimension matched filter is obtained. The physical coordinates of all transmitting and receiving antenna elements participating in the array element position imaging system in space are also obtained. Based on the motion scan-range dimension focusing signal and the array dimension matched filter, the array-range dimension focusing signal is obtained; On a preset 3D imaging grid, the array-distance dimension focusing signal corresponding to each spatial point is acquired to obtain a 3D imaging image of the object to be detected, which is used for subsequent target detection and recognition.

[0005] Based on the above technical solutions, this paper proposes a complete signal processing workflow to address the three core challenges faced by microwave and millimeter-wave 3D imaging systems in complex industrial or security inspection scenarios: strong background clutter interference, focusing misalignment in motion and array dimensions, and distance-array resolution mismatch. These problems are key bottlenecks restricting imaging quality. Without background suppression, weak targets will be obscured; without accurate phase compensation, multi-channel echoes cannot be coherently superimposed; and with sampling mismatch, spatial resolution is limited, resulting in blurred images. To address these issues, this application adopts a three-stage progressive architecture of "background cancellation → motion-dimensional focusing → array-dimensional focusing": accurately subtracting static background in the range domain and windowing to suppress multipath interference; then constructing a motion-frequency and array-dimensional matched filter based on geometric slant range to achieve point-by-point phase correction and coherent accumulation; and simultaneously introducing adaptive interpolation upsampling to match the range dimension sampling density with the array resolution. This solution requires no complex modeling or additional hardware, is entirely driven by a physical propagation model, and is compatible with various antenna configurations such as sparse, curved, and polygonal antennas. It can also output high-fidelity complex images, significantly improving signal-to-noise ratio, focusing accuracy, and target discernibility.

[0006] In conjunction with the first aspect above, in one possible implementation, the method for obtaining the background cancellation echo signal includes: The echo signal of the object to be detected and background echo signal Perform inverse fast Fourier transform (IFFT) on each signal to obtain the range pulse compression signal of the echo signal. Distance pulse compression signal and background echo signal ;in, For the coordinates of the transmitting array element, To receive the coordinates of the array elements, For the coordinates of the motion scan dimension, Wavenumber is the frequency dimension. These are the actual distance dimension coordinates; The range pulse compression signal based on the echo signal and the background echo signal is obtained through the formula. Calculated background pulse pressure signal ; Background pulse pressure signal cancellation A rectangular window is added to the distance dimension, and a Fast Fourier Transform (FFT) is performed to obtain the background cancellation echo signal. .

[0007] In conjunction with the first aspect above, in one possible implementation, the rectangular window includes: The center position of the rectangular window is the distance from the center of the imaging field of view. Rectangular window width The rectangular window function is determined based on the distance distribution range of the object to be detected. Represented as: .

[0008] In conjunction with the first aspect above, in one possible implementation, the method for determining the maximum detection distance in the distance dimension includes: Based on the imaging system's transmission frequency The bandwidth was calculated. ;in, The starting frequency of the transmitted signal, The termination frequency of the transmitted signal; Distance dimension resolution is obtained based on bandwidth calculation. And through the formula The maximum detection range was calculated. ;in, The number of frequency sweep points in the frequency dimension. It is the speed of light.

[0009] In conjunction with the first aspect above, in one possible implementation, the method for obtaining the motion-frequency dimension space matched filter includes: Based on the distances corresponding to each distance grid point and the motion scan position, using the formula... The motion-frequency dimension space matched filter is calculated. ;in, , The frequency of the signal transmitted by the microwave millimeter-wave imaging system; The grid coordinates are divided by the distance dimension, with a maximum value of .

[0010] In conjunction with the first aspect above, in one possible implementation, the method for acquiring the motion scanning-distance dimension focusing signal includes: The background cancellation echo signal is convolved with the motion-frequency dimension spatial matched filter in the motion scan dimension in the time domain to obtain the time-domain convolution signal. ; ; in, In order to be in Temporal convolution operations of dimension 1, background cancellation echo signal before performing convolution operations. Matched filter with motion-frequency dimension of Dimensions zeroed out , Number of sampling points in the motion scan dimension; These are the coordinates of the motion dimension image obtained after temporal convolution. The temporal convolution signal is integrated and accumulated in the frequency dimension to obtain the motion scan-range dimension focusing signal. ; ; in, The spatial frequency wavenumber sampling interval, For the first The wave number corresponding to each frequency .

[0011] In conjunction with the first aspect above, in one possible implementation, the method for obtaining the array dimension matched filter includes: Calculate the emission slant range from each element position to each imaging grid point: ;in, The coordinates of the transmitting array elements; Let these be the coordinates of the imaging grid points in the horizontal dimension. The coordinates of the imaging grid points in the distance dimension; Calculate the receiving slant range from each array element position to each imaging grid point: ;in, The coordinates of the receiving array element; Based on the transmit slant range and receive slant range from each array element position to each imaging grid point, the array dimension matched filter is calculated. ; ; in, The minimum wavenumber in the frequency dimension. .

[0012] In conjunction with the first aspect above, in one possible implementation, the method for acquiring the array-distance dimension focusing signal includes: The distance dimension of the motion scan-distance dimension focusing signal is interpolated and upsampled to obtain the upsampled signal; The upsampled signal is indexed in the distance dimension to obtain an index signal; the array dimension matched filter performs a matched filtering operation on the index signal in the array dimension to obtain an array dimension matched filtered signal. The array-dimensional matched filter signal is integrated and accumulated along the array dimension to obtain the array-distance dimension focused signal.

[0013] In conjunction with the first aspect above, in one possible implementation, determining the sampling factor of the interpolation upsampling includes: Through formula The resolution of the array dimension is calculated. ;in, The center wavelength of the transmitted signal. , The receiving antenna has a 3dB beamwidth. If distance dimension resolution Upsampling factor N=1; if the distance dimension resolution Upsampling factor ;in, This indicates rounding up to the nearest integer.

[0014] Secondly, this application provides a sparse array temporal convolutional joint backprojection imaging system, comprising: an acquisition module, a processing module, and an imaging module; wherein, the acquisition module is used to acquire the backscattered echo signal of the object to be detected, and perform background cancellation processing with the pre-acquired background echo signal to obtain a background cancellation echo signal; the processing module is used to determine the maximum detection distance in the distance dimension according to the bandwidth and number of frequency points of the imaging system, and acquire a distance grid based on the maximum detection distance; calculate a motion-frequency dimension spatial matching filter based on the distance and motion scan dimension coordinates corresponding to each distance grid point; acquire a motion scan-distance dimension focusing signal based on the background cancellation echo signal and the motion-frequency dimension spatial matching filter; acquire an imaging grid in the array dimension and distance dimension, analyze the transmit slant range and receive slant range from each array element position to each imaging grid point, and calculate an array dimension matching filter; acquire an array-distance dimension focusing signal based on the motion scan-distance dimension focusing signal and the array dimension matching filter; the imaging module is used to acquire the array-distance dimension focusing signal corresponding to each spatial point on a preset three-dimensional imaging grid to obtain a three-dimensional imaging image of the object to be detected.

[0015] This application provides a sparse array temporal convolutional combined with back projection imaging method and system, which can achieve high-resolution, high signal-to-noise ratio 3D radar imaging while reducing hardware costs and data acquisition complexity. This method effectively suppresses environmental clutter and system-fixed interference through background cancellation, significantly improving weak target detection capabilities. By combining temporal convolution with back projection, matched filters are constructed in both the motion-frequency and array-range dimensions to achieve staged focusing: pre-focusing is performed in the scanning motion and range dimensions, and then combined with the sparse array geometry to complete fine spatial imaging. This retains the adaptability of the back projection algorithm to arbitrary array configurations and broadband signals while significantly reducing the computational burden of traditional full 3D back projection. Simultaneously, the transmission and reception paths are accurately modeled based on the actual slant range to ensure imaging geometric accuracy. This scheme is particularly suitable for scenarios with limited array element numbers and complex platform motion, such as portable security checks, balancing computational efficiency, system robustness, and engineering feasibility while ensuring imaging quality.

[0016] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0017] Figure 1 A system architecture diagram of a sparse array temporal convolutional joint back projection imaging system provided in this application embodiment; Figure 2 A schematic flowchart of a sparse array temporal convolution joint back projection imaging method provided in an embodiment of this application; Figure 3 A schematic flowchart of a background cancellation processing method provided in an embodiment of this application; Figure 4 A schematic diagram of a microwave millimeter-wave sparse linear array element arrangement provided for an embodiment of this application; Figure 5 This is a schematic diagram of the imaging result of a microwave millimeter-wave sparse linear array on a point target, provided as an embodiment of this application. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The sparse array temporal convolution joint back projection imaging method provided in this application embodiment can be applied to a sparse array temporal convolution joint back projection imaging system, such as... Figure 1 As shown, the communication system includes: an acquisition module, a processing module, and an imaging module; The acquisition module is used to acquire the backscattered echo signal of the object to be detected and perform background cancellation processing with the pre-acquired background echo signal to obtain the background cancellation echo signal. The processing module is used to determine the maximum detection range in the range dimension based on the bandwidth and number of frequency points of the imaging system, obtain the range grid based on the maximum detection range, and obtain the motion-frequency dimension spatial matching filter based on the distance and motion scan dimension coordinates corresponding to each range grid point. Based on the background cancellation echo signal and the motion-frequency dimension spatial matching filter, the motion scanning-range dimension focusing signal is obtained; Imaging grids are obtained in the array dimension and distance dimension. The transmit slant range and receive slant range from each array element position to each imaging grid point are analyzed, and the array dimension matched filter is obtained. Based on the motion scan-range dimension focusing signal and the array dimension matched filter, the array-range dimension focusing signal is obtained; The imaging module is used to acquire the array-distance dimension focusing signal corresponding to each spatial point on a preset three-dimensional imaging grid to obtain a three-dimensional imaging image of the object to be detected.

[0020] To address the technical issues of imaging accuracy problems in microwave and millimeter-wave 3D imaging systems caused by inaccurate system modeling, multidimensional signal mismatch, and strong clutter interference, this application provides a sparse array temporal convolutional joint backprojection imaging method. This method includes: acquiring the backscattered echo signal of the object to be detected and performing background cancellation processing with a pre-acquired background echo signal to obtain a background cancellation echo signal; determining the maximum detection distance in the distance dimension based on the bandwidth and number of frequency points of the imaging system, and obtaining a distance grid based on the maximum detection distance; obtaining a motion-frequency dimension spatial matching filter based on the distance and motion scan dimension coordinates corresponding to each distance grid point; obtaining a motion scan-distance dimension focusing signal based on the background cancellation echo signal and the motion-frequency dimension spatial matching filter; obtaining an imaging grid in the array dimension and distance dimension, analyzing the transmit slant range and receive slant range from each array element position to each imaging grid point, and obtaining an array dimension matching filter; and finally, based on the motion scan-distance dimension focusing signal and the array... The method employs a column-dimensional matched filter to acquire the array-range dimension focusing signal. On a pre-defined 3D imaging grid, the array-range dimension focusing signal corresponding to each spatial point is acquired, resulting in a 3D imaging image of the object to be detected. Based on this, background cancellation processing is introduced to effectively suppress environmental clutter and system fixed spurious responses, significantly improving the signal-to-clutter ratio and enhancing the detection capability for weak targets. The constructed motion-frequency dimension spatial matched filter and array-dimensional matched filter achieve precise focusing in the range-motion scanning domain and array-range domain, respectively, taking into account the scanning characteristics of the motion platform and the spatial sampling structure of the sparse array, overcoming the problems of high computational cost and blurred focusing in traditional back projection algorithms. This method organically combines temporal convolution and back projection, retaining the high-precision imaging advantages of back projection while significantly reducing the computational complexity of subsequent 3D imaging through matched filtering pre-focusing. The entire process is naturally adapted to sparse arrays and wideband systems, reducing hardware costs and data acquisition burden while ensuring imaging resolution. In summary, this method achieves efficient, high-resolution, and low-complexity 3D radar imaging in complex backgrounds.

[0021] like Figure 2 As shown in the embodiment of this application, a sparse array temporal convolutional joint backprojection imaging method is provided, comprising: S201. Acquire the backscattered echo signal of the object to be detected, and perform background cancellation processing with the pre-acquired background echo signal to obtain the background cancellation echo signal.

[0022] S202. Determine the maximum detection distance in the distance dimension based on the bandwidth and number of frequency points of the imaging system, and obtain the distance grid based on the maximum detection distance; obtain the motion-frequency dimension spatial matching filter based on the distance corresponding to each distance grid point and the motion scanning position. The imaging system is a microwave millimeter-wave imaging system.

[0023] S203. Based on the background cancellation echo signal and the motion-frequency dimension spatial matching filter, obtain the motion scanning-distance dimension focusing signal.

[0024] S204. Obtain the imaging grid in the array dimension and the distance dimension, analyze the transmit slant range and receive slant range from each array element position to each imaging grid point, and obtain the array dimension matched filter. Among them, the physical coordinates of all transmitting and receiving antenna elements participating in the array element position imaging system in space.

[0025] S205. Based on the motion scan-distance dimension focusing signal and the array dimension matching filter, obtain the array-distance dimension focusing signal.

[0026] S206. On the preset three-dimensional imaging grid, acquire the array-distance dimension focusing signal corresponding to each spatial point to obtain a three-dimensional imaging image of the object to be detected, which is used for subsequent target detection and recognition.

[0027] Specifically, after traversing all the imaging grids pixel by pixel, the final three-dimensional complex image is obtained; The modulus of the three-dimensional complex image is calculated to obtain a three-dimensional amplitude image, which is then input into three-dimensional target detection and recognition software for subsequent target detection and classification, and finally transmitted to the display terminal for image display.

[0028] Based on the above technical solutions, a sparse array element arrangement method and array configuration are proposed for arc-shaped, polygonal, and linear sparse arrays, respectively. The main methods include background cancellation, temporal convolution, range integral accumulation, and back projection. Background cancellation is achieved by pulse compression processing of target echo data and background echo data, converting them to the range domain for subtraction, and then converting them to the frequency domain for subsequent processing. This can effectively eliminate the influence of signal coupling between the transmitting and receiving antennas and multipath clutter caused by multipath effects, and effectively improve the dynamic range of the resulting image. Temporal convolution achieves focusing of the motion scan dimension image, while distance integral accumulation achieves focusing of the distance dimension image, thus realizing focusing of the motion scan-distance dimension image. Subsequently, a back projection module is used to focus the array dimension and distance dimension images, resulting in a three-dimensional focused complex image for subsequent detection, recognition, and classification processing. This application has high algorithm parallelism, thus achieving high imaging computation efficiency. At the same time, the processing of echo data is all in the time domain, and the proposed method has high imaging accuracy. It can not only be used in the field of non-destructive testing of missing products inside bags on production lines, but also in the fields of automotive radar MIMO imaging and human security inspection imaging technology. It is a good method worthy of promotion.

[0029] In one possible implementation of the embodiments of this application, such as Figure 3As shown, the above S201 can be specifically implemented through the following S301, S302, S303 and S304, which are explained in detail below: S301, The echo signal of the object to be detected and background echo signal Perform inverse fast Fourier transform (IFFT) on each signal to obtain the range pulse compression signal of the echo signal. Distance pulse compression signal and background echo signal ;in, For the coordinates of the transmitting array element, To receive the coordinates of the array elements, For the coordinates of the motion scan dimension, Wavenumber is the frequency dimension. These are the actual distance dimension coordinates; in, This represents the actual discrete sampling point location obtained in the distance dimension after performing an IFFT on the echo signal; it is a variable in the data domain. Its value is determined by the system bandwidth and the number of frequency points, representing the distribution of echo energy in the distance direction.

[0030] It should be noted that performing IFFT on each spatial location (transmitting element, receiving element, moving scan position) is equivalent to converting frequency information into time information.

[0031] S302. The range pulse compression signal based on the echo signal and the background echo signal is obtained through the formula... Calculated background pulse pressure signal ; It should be noted that in microwave millimeter-wave imaging systems, the system transmits a continuously frequency-converted (FMCW) signal and receives reflected echoes from objects. The echo signal consists of two parts: the target signal, which is the object of interest to be detected (such as cigarette packs inside a box); and the background signal, which is fixed clutter around the system (such as conveyor belts, racks, and walls). The target and background are superimposed, and the background must be removed first in order to see the target clearly.

[0032] For example, the antenna array configuration of the microwave and millimeter-wave three-dimensional imaging system of this application can be arc-shaped, polygonal, or straight. The array elements are arranged in a sparse manner. The switching between transmitting and receiving array elements is controlled by a timing controller. Each transmitting array element corresponds to eight receiving array elements. When all eight receiving array elements are activated simultaneously, the multi-channel acquisition board acquires echo data. The frequency range of the transmitted signal from the microwave and millimeter-wave transceiver front end is [missing information]. It adopts either stepped continuous wave (SFCW) or linear frequency modulated continuous wave (FMCW) waveform modes, and the number of sweep points is set to... The dimensions of the acquired echo data are , The number of transmitting array elements, For the number of receiving array elements, The number of sampling points in the motion scan dimension. This represents the number of frequency sweep points in the frequency dimension.

[0033] S303, background pulse pressure signal cancellation The distance dimension is used to add a rectangular window to suppress multipath clutter interference signals; where the center position of the rectangular window is the distance from the center of the imaging field of view. Rectangular window width Based on the distance distribution range of the object to be detected, the rectangular window function is expressed as: ; It should be noted that multipath clutter interference signals: electromagnetic waves may be reflected multiple times by the ground and walls before reaching the receiver, causing false echoes; Add a rectangular window: retain the distance range within the region of interest and ignore irrelevant signals at a distance; for example, only focus on targets between 0.8 and 1.2 meters, and set all others to zero.

[0034] S304. Perform a Fast Fourier Transform (FFT) on the windowed background cancellation pulse signal to obtain the background cancellation echo signal. .

[0035] It should be noted that S303 has obtained a clean signal in the range domain (background has been eliminated and windowed); now it is converted back to the frequency domain for subsequent focusing processing of the motion scanning dimension and the array dimension; that is, from "range information" back to "frequency information", in preparation for the next step of matched filtering.

[0036] Based on the above technical solutions, in microwave and millimeter-wave 3D imaging systems, the echo signal of the target to be detected (such as contraband in a box) is usually overwhelmed by strong background clutter (such as conveyor belts, metal supports, walls, and other fixed structures), resulting in an extremely low signal-to-noise ratio and making it difficult to effectively extract target features. Without background suppression, subsequent imaging will contain a large number of false scattering points, seriously affecting the accuracy of target detection and recognition. Therefore, it is necessary to solve the key technical problem of "weak target signal extraction under strong background interference". This application adopts a processing flow of "frequency domain → range domain → background cancellation → windowed filtering → return to frequency domain", which has significant advantages: the echo is converted to the range domain by IFFT, so that the target and the background are initially separated in space; accurate background cancellation is achieved by direct subtraction in the range domain, with clear physical meaning and high computational efficiency; the introduction of a rectangular window can effectively suppress multipath clutter (such as secondary reflections from the ground / wall), retaining only the signal of the effective area within the field of view, thus improving the signal-to-noise ratio; the FFT restores the signal to the frequency domain, which not only retains the phase information of the original signal, but also is compatible with subsequent frequency domain-based matched filtering and back projection focusing algorithms. This approach requires no complex modeling, has low computational overhead, and can fully preserve the complex characteristics of the target, providing high-quality input for high-resolution 3D imaging and significantly improving the robustness and practicality of the system in complex industrial or security inspection scenarios.

[0037] In one possible implementation of this application embodiment, the above-mentioned S202 can be specifically implemented by the following S401, S402 and S403, which are described in detail below: S401, Based on the transmission frequency of the imaging system The bandwidth was calculated. ;in, The starting frequency of the transmitted signal, The termination frequency of the transmitted signal; Distance dimension resolution is obtained based on bandwidth calculation. And through the formula The maximum detection range was calculated. .

[0038] in, For speed of light; for frequency sweep points. The system determines the maximum unambiguous distance that can be covered; system data acquisition. Each frequency point (e.g., from 10GHz to 11GHz, a total of 1001 points) corresponds to an "echo time" or "slant range". The maximum detection distance is determined by the last frequency point.

[0039] S402, Maximum detection distance based on distance dimension Divide the distance grid .

[0040] S403. Based on the distance and motion scan dimension coordinates corresponding to each distance grid point, using the formula... The motion-frequency dimension space matched filter is calculated. ;in, , The frequency at which the signal is transmitted by the microwave millimeter-wave imaging system.

[0041] Based on the above technical solutions, in microwave and millimeter-wave imaging systems, to achieve high-resolution three-dimensional imaging, it is essential to precisely control and calculate the system's bandwidth, range resolution, and maximum detection range. These parameters directly affect the quality of the final image and the system's performance. Therefore, solving the problem of accurately calculating bandwidth, resolution, and maximum detection range based on the transmission frequency, and performing effective signal processing accordingly, is a key technical issue for improving imaging quality. In S401, by calculating the bandwidth and corresponding range resolution, the minimum range interval (i.e., resolution) that the system can resolve is determined, which is the foundation for ensuring imaging accuracy. Furthermore, the maximum detection range calculated by the formula ensures that the system can cover the desired detection depth without blurring. The importance of this step lies in the fact that it not only defines the effective working range of the system but also provides a basis for subsequent range grid division. In S402, based on the calculated maximum detection range, the range dimension is divided into several grid points. This discretization method facilitates independent signal processing operations on each grid point, thereby improving the efficiency and accuracy of data processing. Reasonable grid division ensures uniform and fine sampling throughout the entire detection range, which is crucial for improving imaging resolution. In S403, a motion-frequency dimension spatial matched filter is calculated based on the distance and motion scan coordinates corresponding to each distance grid point. The core of this step lies in designing a specific phase compensation function for each detection position to eliminate echo signal phase differences caused by the object's different positions relative to the antenna array. This method effectively achieves target focusing, enhances the echo intensity in the target area, and suppresses interference from non-target areas, greatly improving imaging contrast and clarity.

[0042] In summary, this technical solution significantly improves the resolution and anti-interference capability of the imaging system by accurately calculating bandwidth, resolution, and detection distance, rationally dividing the distance grid, and designing a specific spatial matching filter for each grid point. Its advantages lie in ensuring high imaging accuracy while effectively handling complex environmental noise, making it suitable for various application scenarios such as security inspection and medical imaging, and possessing broad practical value.

[0043] In one possible implementation of this application embodiment, the above-mentioned S203 can be specifically implemented by the following S501 and S502, which are described in detail below: S501. Perform a time-domain convolution operation between the background cancellation echo signal and the motion-frequency dimension spatial matched filter in the motion scan dimension to obtain the time-domain convolution signal. : ; in, In order to be in Temporal convolution operations of dimension 1, background cancellation echo signal before performing convolution operations. Matched filter with motion-frequency dimension of Dimensions zeroed out To avoid the circular convolution effect, Number of sampling points in the motion scan dimension; These are the coordinates of the motion dimension image obtained after temporal convolution. Convolution results show that at a fixed distance Below, the superposition response of echoes from all moving positions after phase correction.

[0044] S502. Integrate and accumulate the time-domain convolutional signal in the frequency dimension to obtain the motion scan-distance dimension focused signal. The specific implementation method is as follows: ; in, The spatial frequency wavenumber sampling interval, For the first The wave number corresponding to each frequency .

[0045] It should be noted that this step achieves focusing of the motion direction; for a real target, the echoes it generates at different motion positions will experience phase shifts due to changes in path length; motion-frequency dimension space matched filter This precisely compensates for these phase shifts in the reverse direction; after convolution, all echoes from the same target will be... The coherent accumulation at the point of intersection forms a strong peak.

[0046] Based on the above technical solutions, in microwave and millimeter-wave 3D imaging, the system typically moves the antenna array along the motion scanning dimension (such as the direction of a conveyor belt) mechanically or electronically to obtain multi-angle echo data. However, since the relative position between the target and the antenna changes continuously with scanning, the echoes generated by the same target at different scanning positions will introduce significant phase shifts due to differences in propagation paths. If not corrected, these phase mismatches will cause the signals to cancel each other out during synthesis, severely reducing imaging focus and signal-to-noise ratio, and even causing weak targets to be completely submerged in noise. Therefore, it is urgent to solve the key technical problem of "how to achieve phase consistency compensation during motion scanning to achieve effective focusing". This solution adopts a two-step focusing strategy of "temporal convolution + frequency integration": in S501, the echo signal after background cancellation is linearly convolved with a pre-constructed motion-frequency dimension matched filter in the motion dimension with zero-filling to accurately compensate for the path phase difference at each scanning position; in S502, the convolution result is integrated and accumulated in the frequency dimension to achieve full-band energy convergence. This method is essentially a coherent focusing process based on the principle of back projection. Its advantages include: accurate physical model—the matched filter is strictly constructed according to the geometric slant range, ensuring accurate phase compensation; strong anti-interference capability—only real targets can achieve full-path coherent superposition, effectively suppressing clutter; strong compatibility—applicable to arbitrary scanning trajectories and sparse array configurations; and preservation of complex information—providing complete amplitude and phase features for subsequent array-dimensional focusing and target recognition. In summary, this technical solution achieves high-precision motion-dimensional focusing with low computational complexity, making it an indispensable key step in constructing high-quality three-dimensional complex images.

[0047] In one possible implementation of this application embodiment, the above-mentioned S204 can be specifically implemented by the following S601, S602 and S603, which are described in detail below: S601. Calculate the emission slant range from the position of each array element to each imaging grid point. ;in, Let these be the coordinates of the imaging grid points in the horizontal dimension. The coordinates of the imaging grid points in the distance dimension; It should be pointed out that, This represents the coordinates of the theoretical target location in the distance dimension within the preset 3D imaging space used to reconstruct the image; it is a variable in the reconstruction domain. Its value is determined by the user-defined imaging field of view and grid resolution, and is used to guide matched filtering and back projection calculations.

[0048] S602. Calculate the receiving slant distance from the position of each array element to each imaging grid point. ; Based on the transmit slant range and receive slant range from each array element position to each imaging grid point, the array dimension matched filter is calculated. : ; in, The minimum wavenumber in the frequency dimension. .

[0049] In one embodiment, the divided imaging grid They are respectively represented as, The grid spacing is 1. Therefore, the number of grid points in the divided array dimension is ; The grid spacing is 1. Therefore, the number of grid points in the distance dimension is .

[0050] Based on the above technical solutions, in microwave and millimeter-wave 3D imaging systems, antenna arrays often employ sparse arrangements to reduce hardware costs and system complexity. However, the resulting insufficient spatial sampling can lead to phase mismatch in multi-channel echoes from the same target due to differences in propagation paths, making coherent superposition difficult and causing image blurring, increased sidelobes, or even target misses. To address this, this solution constructs an array dimension matched filter based on a precise geometric model using S601–S602: calculating the transmit slant range and receive slant range from each transmitting and receiving element to the 3D imaging grid point, and calculating the array dimension matched filter accordingly. Matched filtering and coherent accumulation are then performed point-by-point on a pre-defined regular grid to achieve high-precision back projection imaging. This method strictly follows the physical laws of electromagnetic wave propagation, requires no beamforming assumptions or interpolation approximations, and is suitable for various sparse array configurations such as straight lines, arcs, and polygonal lines. It can also output complete complex images to support subsequent phase-sensitive analysis, significantly improving imaging resolution and target detection reliability.

[0051] In one possible implementation of this application embodiment, the above-mentioned S205 can be specifically implemented by the following S701, S702 and S703, which are described in detail below: S701. Interpolate and upsample the distance dimension of the motion scan-distance dimension focusing signal to obtain the upsampled signal; The specific method for determining the sampling factor of the interpolation upsampling is as follows: Through formula The resolution of the array dimension is calculated. ;in, The center wavelength of the transmitted signal. , The receiving antenna has a 3dB beamwidth. If distance dimension resolution This indicates that the distance resolution is already good enough and upsampling is not necessary; the upsampling factor N=1. If the distance dimension resolution... Upsampling factor ;in, This indicates rounding up to the nearest integer.

[0052] In one embodiment, the receiving antenna beamwidth is designed to be The center wavelength of the transmitted signal is The resulting array dimension resolution is Therefore, the upsampling factor is N=5, and the number of points after upsampling is .

[0053] It should be noted that when the distance dimension resolution is higher than the array dimension (i.e., This can lead to "insufficient spatial sampling," making it impossible to fully utilize array information; interpolation upsampling can increase the number of sampling points in the distance dimension, making subsequent matched filtering more accurate; similar to "enlarging a low-resolution image to a high-resolution one" in order to better capture details.

[0054] S702. Index the sampling points of the upsampled signal in the distance dimension to obtain the index signal; perform matched filtering on the index signal in the array dimension using the array dimension matched filter to obtain the array dimension matched filtered signal. S703. Integrate and accumulate the array-dimension matched filter signal along the array dimension to obtain the array-distance dimension focused signal.

[0055] Based on the above technical solutions, in microwave and millimeter-wave 3D imaging systems, the range dimension resolution is determined by the signal bandwidth, while the array dimension resolution is limited by the antenna beamwidth and physical aperture. When the range resolution (e.g., 15cm) is significantly better than the array resolution (e.g., 2.8cm), directly using the original range sample for focusing will result in insufficient spatial sampling in the array direction, making it impossible for matched filtering to accurately align with the target position, causing energy diffusion, image blurring, and positioning errors. To solve this "dimensional resolution mismatch" problem, this solution introduces an interpolation upsampling strategy based on the ratio of array to range resolution in S701: by calculating... and The method adaptively determines the upsampling factor N based on the relationship between the sampling density and the range, and performs range-dimensional interpolation on the motion-range focusing signal to match the array dimension resolution. In S702–S703, range indexing, array-dimensional matched filtering, and coherent accumulation are combined with the geometric slant range model to achieve high-precision focusing. This method does not require increasing hardware bandwidth or the number of antennas; it improves spatial consistency solely through signal domain resampling, significantly enhancing the sharpness of the point spread function and the imaging signal-to-noise ratio. It is also compatible with sparse arrays and arbitrary scanning configurations, achieving array-range joint high-resolution imaging at a low computational cost, providing a high-quality complex image foundation for subsequent target detection and recognition.

[0056] This invention targets the field of microwave and millimeter-wave three-dimensional imaging technology. Compared to existing X-ray imaging technology, microwave and millimeter-wave imaging technology emits no ionizing radiation, posing no health threat to operators, and achieves millimeter-level resolution, ensuring detection effectiveness and quality. Compared to existing microwave and millimeter-wave imaging methods, the microwave and millimeter-wave imaging array configuration used in this application can be designed as arc-shaped, polygonal, and linear configurations, adapting to different array configurations. Simultaneously, the antenna elements are arranged in a sparse manner, effectively reducing RF hardware costs. Furthermore, the method in this application performs dimensional decomposition in the echo data dimension, effectively improving algorithm parallelism and ensuring imaging accuracy. The imaging method employed in this invention can adapt to different array configurations, has high algorithm execution efficiency, and possesses high imaging accuracy, making it a promising method worthy of widespread adoption.

[0057] like Figure 4 The diagram shows a microwave / millimeter-wave sparse linear array element arrangement, with the number of transmitting elements being [number missing]. The number of receiving array elements is The spacing between adjacent receiving elements is 7.4 mm, the spacing between adjacent transmitting elements is 29.6 mm, and the equivalent array element spacing is 3.7 mm, satisfying the Nyquist sampling theorem. The equivalent array length is 943.5 mm, satisfying the imaging field of view requirements. Figure 5 The image shown is the imaging result of a point target using a microwave millimeter-wave sparse linear array.

[0058] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

Claims

1. A sparse array temporal convolutional joint backprojection imaging method, characterized in that, include: The backscattered echo signal of the object to be detected is acquired and then background cancellation processing is performed with the pre-acquired background echo signal to obtain the background cancellation echo signal. The maximum detection distance in the distance dimension is determined based on the bandwidth and number of frequency points of the imaging system, and a distance grid is obtained based on the maximum detection distance. A motion-frequency dimension spatial matching filter is obtained based on the distance and motion scan dimension coordinates corresponding to each distance grid point. The distance dimension is a one-dimensional spatial coordinate axis representing the distance between the object to be detected and the antenna along the electromagnetic wave propagation direction. Based on the background cancellation echo signal and the motion-frequency dimension spatial matching filter, the motion scanning-distance dimension focusing signal is obtained; Imaging grids are obtained in the array dimension and distance dimension. The transmit slant range and receive slant range from each array element position to each imaging grid point are analyzed, and the array dimension matched filter is obtained. Based on the motion scan-range dimension focusing signal and the array dimension matched filter, the array-range dimension focusing signal is obtained; On a preset 3D imaging grid, the array-distance dimension focusing signal corresponding to each spatial point is acquired to obtain a 3D imaging image of the object to be detected.

2. The sparse array temporal convolution joint back projection imaging method according to claim 1, characterized in that, The method for obtaining the background cancellation echo signal includes: The echo signal of the object to be detected and background echo signal Perform inverse fast Fourier transform (IFFT) on each signal to obtain the range pulse compression signal of the echo signal. Distance pulse compression signal and background echo signal ;in, For the coordinates of the transmitting array element, To receive the coordinates of the array elements, For the coordinates of the motion scan dimension, Wavenumber is the frequency dimension. These are the actual distance dimension coordinates; The range pulse compression signal based on the echo signal and the background echo signal is obtained through the formula. Calculated background pulse pressure signal ; Background pulse pressure signal cancellation A rectangular window is added to the distance dimension, and a Fast Fourier Transform (FFT) is performed to obtain the background cancellation echo signal. .

3. The sparse array temporal convolution joint back projection imaging method according to claim 2, characterized in that, The rectangular window includes: The center position of the rectangular window is the distance from the center of the imaging field of view. Rectangular window width The rectangular window function is determined based on the distance distribution range of the object to be detected. Represented as: 。 4. The sparse array temporal convolution joint back projection imaging method according to claim 1, characterized in that, The method for determining the maximum detection range in the distance dimension includes: Based on the imaging system's transmission frequency The bandwidth was calculated. ;in, The starting frequency of the transmitted signal, The termination frequency of the transmitted signal; Distance dimension resolution is obtained based on bandwidth calculation. And through formula The maximum detection range was calculated. ;in, The number of frequency sweep points in the frequency dimension. It is the speed of light.

5. The sparse array temporal convolution joint back projection imaging method according to claim 2, characterized in that, The method for obtaining the motion-frequency dimension space matched filter includes: Based on the distances corresponding to each distance grid point and the motion scan position, using the formula... The motion-frequency dimension space matched filter is calculated. ;in, , The frequency of the signal transmitted by the microwave millimeter-wave imaging system; The grid coordinates are divided by the distance dimension, with a maximum value of .

6. The sparse array temporal convolution joint back projection imaging method according to claim 2, characterized in that, The method for acquiring the motion scanning-range dimension focusing signal includes: The background cancellation echo signal is convolved with the motion-frequency dimension spatial matched filter in the motion scan dimension in the time domain to obtain the time-domain convolution signal. ; ； in, In order to be in Temporal convolution operations of dimension 1, background cancellation echo signal before performing convolution operations. Matched filter with motion-frequency dimension of Dimensions zeroed out , Number of sampling points in the motion scan dimension; These are the coordinates of the motion dimension image obtained after temporal convolution. The temporal convolution signal is integrated and accumulated in the frequency dimension to obtain the motion scan-range dimension focusing signal. ; ； in, The spatial frequency wavenumber sampling interval, For the first The wave number corresponding to each frequency .

7. The sparse array temporal convolution joint backprojection imaging method according to claim 6, characterized in that, The method for obtaining the array dimension matched filter includes: Calculate the emission slant range from each element position to each imaging grid point: ;in, The coordinates of the transmitting array elements; Let these be the coordinates of the imaging grid points in the horizontal dimension. The coordinates of the imaging grid points in the distance dimension; Calculate the receiving slant range from each array element position to each imaging grid point: ;in, The coordinates of the receiving array element; Based on the transmit slant range and receive slant range from each array element position to each imaging grid point, the array dimension matched filter is calculated. ; ； in, The minimum wavenumber in the frequency dimension. , The starting frequency of the transmitted signal, It is the speed of light.

8. The sparse array temporal convolution joint back projection imaging method according to claim 1, characterized in that, The method for acquiring the array-range dimension focusing signal includes: The distance dimension of the motion scan-distance dimension focusing signal is interpolated and upsampled to obtain the upsampled signal; The upsampled signal is indexed in the distance dimension to obtain an index signal; the array dimension matched filter performs a matched filtering operation on the index signal in the array dimension to obtain an array dimension matched filtered signal. The array-dimensional matched filter signal is integrated and accumulated along the array dimension to obtain the array-distance dimension focused signal.

9. A sparse array temporal convolutional joint backprojection imaging method according to claim 8, characterized in that, The determination of the sampling factor for the interpolation upsampling includes: Through formula The resolution of the array dimension is calculated. ;in, The center wavelength of the transmitted signal. , The receiving antenna has a 3dB beamwidth. If distance dimension resolution Upsampling factor N=1; if the distance dimension resolution Upsampling factor ;in, This indicates rounding up to the nearest integer.

10. A sparse array temporal convolution joint back projection imaging system, operating based on the sparse array temporal convolution joint back projection imaging method according to any one of claims 1-9, characterized in that, It includes an acquisition module, a processing module, and an imaging module; The acquisition module is used to acquire the backscattered echo signal of the object to be detected, and perform background cancellation processing with the pre-acquired background echo signal to obtain the background cancellation echo signal. The processing module is used to determine the maximum detection distance in the distance dimension based on the bandwidth and number of frequency points of the imaging system, obtain the distance grid based on the maximum detection distance, and calculate the motion-frequency dimension spatial matching filter based on the distance and motion scanning dimension coordinates corresponding to each distance grid point. Based on the background cancellation echo signal and the motion-frequency dimension spatial matching filter, the motion scanning-distance dimension focusing signal is obtained; The imaging grid is obtained in the array dimension and the distance dimension. The transmit slant range and receive slant range from each array element position to each imaging grid point are analyzed, and the array dimension matched filter is calculated. Based on the motion scan-range dimension focusing signal and the array dimension matched filter, the array-range dimension focusing signal is obtained; The imaging module is used to acquire the array-distance dimension focusing signal corresponding to each spatial point on a preset three-dimensional imaging grid to obtain a three-dimensional imaging image of the object to be detected.

Images