Non-uniform sampling based millimeter wave MIMO radar imaging method, device and product

Abstract

This disclosure belongs to the field of radar imaging technology, and specifically relates to a millimeter-wave MIMO radar imaging method, device, and product based on non-uniform sampling. The method includes: determining the coordinates of the transmitting antenna, the receiving antenna, and the virtual array element; obtaining the coordinates of the mechanical scanning position in the global imaging coordinate system; superimposing the virtual array element coordinates with the mechanical scanning position coordinates to construct a non-uniform aperture sampling point set; using a non-uniform fast Fourier transform, mapping the echo data on the non-uniform aperture sampling point set to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point; performing near-field phase compensation of the echo data in the wavenumber domain; and reconstructing the radar reflection intensity image of the target area using the compensated echo data in the wavenumber domain. The imaging method proposed in this disclosure can solve the problem of dependence on uniform sampling in traditional fast imaging algorithms, improve the channel utilization of actual MIMO virtual arrays, and reduce amplitude and phase errors caused by interpolation resampling.

Description

Technical Field

[0001] This disclosure belongs to the field of radar imaging technology, and specifically relates to a millimeter-wave MIMO radar imaging method, device and product based on non-uniform sampling. Background Technology

[0002] Existing near-field imaging methods for millimeter-wave MIMO-SAR mainly include the RMA / Ω-K fast imaging method based on uniform arrays, the fast imaging method based on interpolation resampling, and the direct back projection imaging method based on BPA. These methods still have the following shortcomings when facing the non-uniform virtual arrays and non-uniform aperture sampling of actual millimeter-wave MIMO radars: (1) Traditional RMA / Ω-K fast imaging methods rely on uniform sampling and are difficult to apply directly to non-uniform virtual arrays: This type of algorithm has a fast calculation speed and is suitable for large-scale radar image reconstruction, but its premise is that the spatial sampling points need to be located on a regular uniform grid. However, in actual millimeter-wave MIMO radar systems, the physical arrangement of transmitting and receiving antennas is limited by factors such as chip packaging, board-level wiring, cascading method, antenna spacing and hardware structure size. The positions of virtual array elements formed by multiple transmitting antennas and multiple receiving antennas often do not meet the strict uniform arrangement. When the positions of virtual array elements do not meet the uniform spacing, if they are still forcibly used as uniform array data and input into the traditional FFT / RMA algorithm, the spatial spectrum calculation will not match the real sampling coordinates, which will lead to phase compensation error. For near-field millimeter-wave imaging, the target distance is relatively close and the propagation path difference is obvious. The phase is extremely sensitive to spatial position error, which will cause problems such as image defocus, side lobe increase, target outline blur or target position shift. (2) Selecting a uniform subarray will discard effective channels, reducing aperture utilization and imaging performance: In order to meet the requirements of uniform sampling in the traditional RMA / Ω-K algorithm, existing engineering solutions often select a portion of approximately uniformly arranged virtual array elements from the complete non-uniform MIMO virtual array to form a uniform subarray, and only use this uniform subarray for fast imaging. This method will lead to a decrease in effective aperture utilization, and the spatial sampling points in the complete virtual array that could originally participate in imaging will not be fully utilized; the echo energy utilization will decrease, and the target reflection information in the discarded channels cannot participate in coherent accumulation, resulting in a decrease in imaging signal-to-noise ratio; the spatial resolution will decrease, and the array's ability to sample target spatial frequency information will be weakened after the number of virtual array elements is reduced; the hardware resources will not be fully utilized, and the actual radar system already has multiple transmit and receive channels, but due to algorithm limitations, they cannot be fully utilized, resulting in a mismatch between hardware capabilities and imaging algorithms; (3) Interpolating non-uniform sampled data onto a uniform grid introduces additional amplitude and phase errors: Existing methods first interpolate non-uniform data onto a regular uniform grid, and then use traditional FFT / RMA imaging. However, millimeter-wave radar imaging processes complex echo data, which contains both amplitude and phase information. Millimeter-wave signals have short wavelengths, and even small errors in spatial position can cause significant phase changes. Especially in near-field imaging, the propagation path between radar elements and the target point has obvious nonlinearity. If interpolation is performed only on non-uniform sampled points at the data level without strictly combining it with the near-field propagation model, the interpolated data cannot truly represent the physical echo of the original sampled data. In addition, in the interpolation method, improper grid selection will result in the loss of high spatial frequency information or increase the computational load, and inappropriate interpolation kernel selection will cause smoothing over-smoothing, sidelobe increase, or local oscillation, introducing additional modeling and numerical errors.

[0003] (4) BPA algorithm but computational complexity is too high: BPA directly performs phase compensation and coherent accumulation based on the real propagation path between each transmitting antenna, each receiving antenna and each imaging point, which has obvious advantages in terms of geometric adaptability. However, for millimeter-wave near-field high-resolution imaging, the imaging area needs a dense spatial grid; for MIMO radar, the number of channels is large; for broadband FMCW radar, the number of range frequency sampling points is large. The BPA's pixel-by-pixel, channel-by-channel, and frequency-by-frequency point accumulation calculation method results in an excessively long imaging time, which is difficult to meet the requirements of fast display, batch processing or near real-time imaging.

[0004] In summary, existing technologies struggle to simultaneously achieve both adaptability to non-uniform arrays and rapid imaging efficiency. Summary of the Invention

[0005] To address the aforementioned problems, this disclosure provides a millimeter-wave MIMO radar imaging method based on non-uniform sampling, which has the advantage of not relying on uniform sampling. The method includes: Determine the coordinates of the transmitting antenna, the receiving antenna, and the virtual array elements; Obtain the coordinates of the mechanical scanning position in the global imaging coordinate system; By superimposing the coordinates of the virtual array elements with the coordinates of the mechanical scanning position, a set of non-uniform aperture sampling points is constructed. A non-uniform fast Fourier transform is used to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point. Near-field phase compensation of echo data in the wavenumber domain; The radar reflection intensity image of the target area is reconstructed from the compensated echo data in the wavenumber domain.

[0006] Furthermore, the virtual array element coordinates are determined by the transmitting antenna coordinates and the receiving antenna coordinates according to the phase center, and all virtual array element positions are arranged at non-equidistant intervals; the mechanical scanning position coordinates are determined according to at least one of the following methods: slide rail motion parameters, trigger timing, or external encoder measurement values.

[0007] Furthermore, the non-uniform aperture sampling point set is composed of MIMO channels and scanning positions; the non-uniformity of the non-uniform aperture sampling point set originates from at least one of the following factors: non-uniform virtual array, mechanical scanning step size error, non-equal interval step size, sparse sampling strategy, missing channels, and irregular array element positions.

[0008] Furthermore, before performing the mapping using a non-uniform fast Fourier transform, the following steps are also included: Equivalent monostatic phase compensation is performed on the MIMO multistatic channel to convert the multistatic propagation distance into the monostatic propagation distance at the virtual array element; the compensation factor of the equivalent monostatic phase compensation is constructed based on the phase difference between the multistatic propagation distance and the monostatic propagation distance at the virtual array element.

[0009] Furthermore, a non-uniform fast Fourier transform is employed to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point. This includes: for each range wavenumber sampling point, calculating the spatial spectrum on the regular wavenumber domain grid based on the real spatial coordinates of each sampling point to achieve the mapping of the echo data.

[0010] Furthermore, the near-field phase compensation is based on a spherical wave propagation model, and the wavenumber domain data is phase-corrected according to the wavenumber in the depth direction; the reconstruction of the radar reflection intensity image of the target area includes: coherently accumulating the focusing results at wavenumber sampling points in different range directions, and obtaining the reflection intensity image of the target area at a specified depth through two-dimensional inverse Fourier transform.

[0011] Furthermore, prior to the non-uniform fast Fourier transform mapping, the method further includes: assigning a weight value to each sampling point based on the distribution density of each sampling point in its neighborhood, and using the weight value to perform amplitude weighting processing on the echo data.

[0012] This disclosure also proposes a millimeter-wave MIMO radar imaging system based on non-uniform sampling, comprising: The coordinate determination module is used to determine the coordinates of the transmitting antenna, the receiving antenna, and the virtual array elements; The coordinate acquisition module is used to acquire the coordinates of the mechanical scanning position in the global imaging coordinate system; The set construction module is used to superimpose the coordinates of virtual array elements with the coordinates of mechanical scanning positions to construct a set of non-uniform aperture sampling points; The data mapping module is used to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point using a non-uniform fast Fourier transform. The data compensation module is used for near-field phase compensation of echo data in the wavenumber domain; The image reconstruction module is used to reconstruct the radar reflection intensity image of the target area from the compensated echo data in the wavenumber domain.

[0013] This disclosure also proposes a millimeter-wave MIMO radar, including a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the above-described millimeter-wave MIMO radar imaging method based on non-uniform sampling.

[0014] This disclosure also proposes a computer program product stored in a computer-readable storage medium, which, when executed by a processor, is used to at least implement the above-described millimeter-wave MIMO radar imaging method based on non-uniform sampling.

[0015] Compared with the prior art, this disclosure has the following advantages: (1) Solving the problem of dependence on uniform sampling in traditional fast imaging algorithms: This disclosure establishes a real spatial coordinate model of the transmitting antenna, receiving antenna, virtual array elements and mechanical scanning positions, and represents radar echo data as non-uniform aperture data located on real spatial sampling points, instead of forcibly assuming that it is located on a regular uniform grid. On this basis, this disclosure uses NUFFT to replace the spatial FFT in the traditional RMA / Ω-K algorithm, so that the complex echo data on non-uniform sampling points can be directly mapped to the regular wavenumber domain grid, thereby breaking through the limitation of uniform spatial sampling in traditional FFT fast imaging methods; (2) Improve the channel utilization of the actual MIMO virtual array: This disclosure does not require the selection of uniform subarrays from the complete virtual array, but rather enables the unified modeling and processing of non-uniform array data based on the coordinates of the actual virtual array elements. Therefore, more effective virtual channels formed by the combination of transmitting and receiving antennas can participate in imaging. By retaining more effective channels, this disclosure can improve the radar aperture utilization and echo energy utilization, reduce information loss caused by channel discarding, and thus help improve the imaging signal-to-noise ratio, spatial resolution, and target detail representation capabilities; (3) Reducing amplitude and phase errors caused by interpolation resampling: This disclosure does not simply interpolate non-uniform spatial sampling data into uniform grid data, but directly inputs the true coordinates of each sampling point into the NUFFT transformation process. This avoids the additional amplitude and phase errors caused by grid selection, interpolation kernel selection and near-field propagation nonlinearity in the traditional interpolation resampling process. Through coordinate-driven non-uniform Fourier transform, this disclosure can more accurately maintain the correspondence between the original non-uniform sampling data and the true spatial position, thereby improving the physical consistency of near-field millimeter-wave imaging; (4) Reduced computational complexity compared to BPA, enabling rapid imaging: This disclosure utilizes NUFFT to rapidly transform non-uniform spatial sampling data into the regular wavenumber domain, and then combines near-field phase compensation, range migration correction, and inverse transformation to obtain the target image. This method avoids the high computational cost of BPA, which calculates the propagation distance for each pixel, each channel, and each frequency sampling point individually and coherently accumulates them. Therefore, this disclosure significantly improves imaging speed while maintaining the adaptability of non-uniform arrays, making it more suitable for rapid imaging applications of large-aperture, high-resolution millimeter-wave MIMO-SAR; (5) Forming a unified imaging framework applicable to non-uniform virtual arrays, non-uniform scanning, and sparse sampling: This disclosure provides a unified description of all aperture sampling points using real spatial coordinates. Regardless of whether the non-uniformity originates from the MIMO virtual array itself, mechanical scanning position errors, sparse sampling strategies, or missing channels, it can all be processed within the same coordinate-driven NUFFT imaging framework. Therefore, the purpose of this disclosure is not only to solve the problem of non-uniform sampling in a single direction, but also to establish a general fast near-field imaging architecture applicable to non-uniform aperture data from millimeter-wave MIMO radar.

[0016] Other features and advantages of this disclosure will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the disclosure. The objects and other advantages of this disclosure may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A millimeter-wave MIMO radar imaging method based on non-uniform sampling according to an embodiment is shown; Figure 2aThe principle of millimeter-wave MIMO radar imaging based on non-uniform sampling according to an embodiment is illustrated; Figure 2b The data processing unit architecture corresponding to millimeter-wave MIMO radar imaging according to an embodiment is shown; Figure 3 The millimeter-wave MIMO radar imaging output results according to an embodiment are shown; Figure 4 The comparison results of a non-uniform virtual array and a uniform array according to the embodiment are shown; Figure 5 The non-uniform aperture sampling point set construction according to an embodiment is shown; Figure 6 This illustrates a NUFFT non-uniform spatial sampling to a regular wavenumber domain grid mapping according to an embodiment; Figure 7 The results of uniform spatial sampling to regular wavenumber domain grid imaging according to an embodiment are shown; Figure 8 The results of 2D imaging reconstruction using sparse array sampling RMA according to an embodiment are shown. Figure 9 The results of 3D imaging reconstruction using sparse array sampling RMA according to an embodiment are shown. Figure 10 The results of 2D imaging reconstruction using sparse array sampling NUFFT-RMA according to an embodiment are shown. Figure 11 The results of 3D imaging reconstruction using sparse array sampling NUFFT-RMA according to an embodiment are shown. Figure 12 The imaging results of non-uniform array spatial sampling according to an embodiment are shown; Figure 13 A millimeter-wave MIMO radar imaging system based on non-uniform sampling according to an embodiment is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0020] like Figure 1 As shown, this disclosure proposes a millimeter-wave MIMO radar imaging method based on non-uniform sampling, which includes: Determine the coordinates of the transmitting antenna, the receiving antenna, and the virtual array elements; Obtain the coordinates of the mechanical scanning position in the global imaging coordinate system; By superimposing the coordinates of the virtual array elements with the coordinates of the mechanical scanning position, a set of non-uniform aperture sampling points is constructed. A non-uniform fast Fourier transform is used to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point. Near-field phase compensation of echo data in the wavenumber domain; The radar reflection intensity image of the target area is reconstructed from the compensated echo data in the wavenumber domain.

[0021] According to some embodiments of this disclosure, such as Figure 2a As shown, the imaging method proposed in this disclosure addresses the problems of non-uniform virtual arrays, non-uniform mechanical scanning sampling, sparse sampling, or missing channels in practical millimeter-wave MIMO radars. By establishing a real spatial sampling coordinate model (including the spatial coordinates of the transmitting antenna, receiving antenna, and virtual array elements, as well as the real coordinate model of the mechanical scanning position), a set of non-uniform aperture sampling points is obtained based on this model. Then, based on the non-uniform fast Fourier transform, the non-uniform aperture sampling data is mapped to a regular wavenumber domain grid for near-field phase compensation and rapid imaging based on the compensated data. Figure 2b In this embodiment, the data processing unit's input is raw ADC data. This data is first sent to the radar data acquisition module, and then sequentially processed by the data reading and rearrangement module, the continuous scan alignment module, the channel amplitude and phase calibration module, the transmit / receive / virtual element coordinate modeling module, and the mechanical scan coordinate modeling module. The data output from the mechanical scan coordinate modeling module is sent to the multi-static to monostatic phase compensation module, and then continues to be processed sequentially by the aperture density weighting module, the NUFFT fast imaging module, the near-field wavenumber domain phase compensation module, the image reconstruction module, and the image display and storage module. The image display and storage module sends control commands and operating status back to the multi-static to monostatic phase compensation module through the control / status feedback path formed by the dashed lines, realizing closed-loop control of the process, and outputting as shown in the figure. Figure 3 The image shown.

[0022] According to some embodiments of this disclosure, such as Figure 2a As shown, the method proposed in this disclosure is based on acquiring raw echo data by placing a MIMO millimeter-wave radar on a two-dimensional moving slide. The principle of echo data acquisition is as follows: the radar system transmits an FMCW frequency-modulated continuous wave signal and collects the target reflected echo through multiple receiving channels. For an FMCW radar, the frequency corresponding to the nth range sampling point can be expressed as:

[0023] in, The starting frequency is K, and the frequency modulation slope is K. This is the ADC start sampling time; The corresponding wave number is:

[0024] in, The speed is the speed of light. The raw echo data obtained in this step includes the MIMO channel dimension, the scan position dimension, and the distance-frequency sampling dimension.

[0025] Furthermore, the virtual array element coordinates are determined by the transmitting antenna coordinates and the receiving antenna coordinates according to the phase center, and all virtual array element positions are arranged at non-equidistant intervals; the mechanical scanning position coordinates are determined according to at least one of the following methods: slide rail motion parameters, trigger timing, or external encoder measurement values.

[0026] According to some embodiments of this disclosure, the spatial coordinates of the virtual array element are determined by the positions of the transmitting antenna and the receiving antenna according to the phase center. The implementation principle is as follows: for the first... The transmitting antenna and the first The virtual array element positions of a receiving antenna can be represented as follows:

[0027] The proposed scheme does not require all virtual array elements to be arranged at equal intervals, but instead retains the arrangement of the actual transmitting antenna. and receiving antenna position The determined real virtual array element coordinates The principle is as follows: For practical cascaded MIMO radars, due to limitations imposed by multiple chips, antenna wiring, antenna spacing, and physical structure, the resulting virtual array elements may not be a regular, uniform, linear array. Traditional RMA / Ω-K algorithms typically require the selection of uniform subarrays, while this disclosure allows the direct use of these non-uniform virtual array element coordinates for subsequent imaging.

[0028] Let the mechanical scanning position be ,in, Indicates the first The scanning position can be determined by the slide rail movement speed, frame period, actual step size, or external encoder measurements. If continuous scanning is used in the horizontal direction, then the 1st scanning position... Horizontal sampling positions It can be represented as:

[0029] in, Indicates the initial scan position; Indicates the actual step size.

[0030] Accordingly, the proposed solution does not require the mechanical scanning positions to be arranged at equal intervals; it is only necessary to record or calculate the actual spatial position corresponding to each sampling point.

[0031] According to some embodiments of this disclosure, such as Figure 4 As shown, the virtual array formed by actual millimeter-wave MIMO radar is usually not an ideal uniform array. However, this invention, through real element coordinate modeling and NUFFT fast spatial spectrum transformation, enables the non-uniform virtual array to still participate in fast near-field imaging, thereby avoiding channel loss caused by selecting uniform subarrays in traditional methods, and also avoiding amplitude and phase errors caused by forcibly interpolating non-uniform data to a uniform grid. Figure 5 As shown, this application constructs a set of real non-uniform aperture sampling points by superimposing virtual array element coordinates and scanning coordinates, which is used for NUFFT fast imaging.

[0032] Furthermore, the non-uniform aperture sampling point set is composed of MIMO channels and scanning positions; the non-uniformity of the non-uniform aperture sampling point set originates from at least one of the following factors: non-uniform virtual array, mechanical scanning step size error, non-equal interval step size, sparse sampling strategy, missing channels, and irregular array element positions.

[0033] According to some embodiments of this disclosure, the virtual element coordinates and mechanical scan coordinates are superimposed to obtain the true aperture position corresponding to each valid echo sample. For the first... The first transmitting antenna, the first The receiving antenna, the first For each scan position, its equivalent sampling position can be represented as:

[0034] in, This indicates the number of a comprehensive sampling point.

[0035] The set of non-uniform aperture sampling points, consisting of all valid channels and all valid scanning positions, is as follows: :

[0036] in, Indicates the first The spatial sampling points at the th Complex echoes at each wavenumber sampling point; Indicates the first spatial sampling points coordinate; Indicates the first spatial sampling points coordinate.

[0037] According to some embodiments of this disclosure, the non-uniformity of the non-uniform aperture sampling point set originates from at least one of the following: non-uniform virtual array, mechanical scanning step size error, non-equidistant step size, sparse sampling strategy, missing channels, and irregular array element positions. Through the above-described superposition method and set construction, the non-uniform virtual array, non-uniform mechanical scanning, sparse scanning, and missing channels can all be uniformly represented as a non-uniform sampling set driven by real coordinates, participating in subsequent imaging processing.

[0038] Furthermore, before performing the mapping using a non-uniform fast Fourier transform, the following steps are also included: Equivalent monostatic phase compensation is performed on the MIMO multistatic channel to convert the multistatic propagation distance into the monostatic propagation distance at the virtual array element; the compensation factor of the equivalent monostatic phase compensation is constructed based on the phase difference between the multistatic propagation distance and the monostatic propagation distance at the virtual array element.

[0039] According to some embodiments of this disclosure, each channel of a real MIMO radar is a multi-static channel, meaning the transmitting antenna and receiving antenna are located at different positions, and the target echo propagation distance is:

[0040] in, Indicates the distance the target echo has traveled; The target spatial location.

[0041] According to some embodiments of this disclosure, in order to make subsequent wavenumber domain rapid imaging closer to the monostatic imaging model, this disclosure can perform equivalent monostatic phase compensation on MIMO multistatic data. Let the virtual element position be... Then the equivalent monobase propagation distance is:

[0042] The compensation factor for equivalent monostatic phase compensation is constructed based on the phase difference between the multistatic propagation distance and the monostatic propagation distance at the virtual array element, and its expression is:

[0043] The compensated echo is:

[0044] After the above compensation, the multi-transmitter, multi-receiver MIMO channel is equivalent to a monostatic observation channel located at the virtual array element, thus organizing the data into a form suitable for NUFFT input after multistatic-to-monostatic compensation. For regular storage format, the data can be organized as follows:

[0045] in, Corresponding horizontal scan coordinates Corresponding to virtual array or composite vertical aperture coordinates, Corresponding to the distance wavenumber. If the vertical direction is formed by both a virtual array and multi-line scanning, then the composite vertical coordinates are:

[0046] This unifies the "virtual array aperture" and the "mechanical scanning aperture" into the same physical coordinate system, enabling the subsequent NUFFT wavenumber domain imaging process to be executed efficiently.

[0047] Furthermore, a non-uniform fast Fourier transform is employed to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the true spatial coordinates of each sampling point, including: For each range wavenumber sampling point, the spatial spectrum on the regular wavenumber domain grid is calculated based on the real spatial coordinates of each sampling point to achieve the mapping of echo data.

[0048] According to some embodiments of this disclosure, the conventional RMA / Ω-K algorithm uses FFT to transform uniform spatial data to the wavenumber domain, i.e. However, FFT requires and The directional sampling points must lie within a regular, uniform grid. However, this condition is often not met for non-uniform virtual arrays or non-uniform scan data. Therefore, this disclosure uses NUFFT instead of traditional spatial FFT. For each range wavenumber... Perform a non-uniform spatial Fourier transform based on the actual sampled coordinates:

[0049] in, Represents data on a regular wavenumber domain grid. Indicates the first At the spatial sampling location and in the range wavenumber Below, the complex radar echo amplitude after multi-base phase compensation is obtained. and These are the transverse and longitudinal wavenumbers, respectively.

[0050] The purpose of this operation is to directly use the real coordinates of non-uniform sampling points to complete fast wavenumber domain mapping, rather than discarding channels or simply interpolating the data to a uniform grid.

[0051] Furthermore, the near-field phase compensation is based on a spherical wave propagation model, and the wavenumber domain data is phase-corrected according to the wavenumber in the depth direction; the reconstruction of the radar reflection intensity image of the target area includes: coherently accumulating the focusing results at wavenumber sampling points in different range directions, and obtaining the reflection intensity image of the target area at a specified depth through two-dimensional inverse Fourier transform.

[0052] According to some embodiments of this disclosure, after NUFFT, the data has been mapped to a regular wavenumber domain grid. To achieve near-field focusing, further calculation of the depth-direction wavenumber is required. :

[0053] Among them, 2 This represents the total number of waves corresponding to round-trip propagation from a single base station.

[0054]

[0055] in, This represents the wavenumber domain complex echo data that achieves coherent focusing after near-field phase compensation; To The compensation factor.

[0056] This operation enables coherent focusing of components with different frequencies and spatial wavenumbers at a specified depth.

[0057] For a specified depth The calculation process for the two-dimensional imaging at that location is as follows: First, the focusing results at all distance-to-wavenumber sampling points are summed, as shown in the following expression:

[0058] Subsequently, the target region was obtained through a two-dimensional inverse Fourier transform. Image of reflection intensity at location :

[0059] The output image can be displayed as a linear amplitude graph or a dB dynamic range graph. Areas with higher brightness in the image represent targets with stronger millimeter-wave reflection at that location.

[0060] According to some embodiments of this disclosure, such as Figure 7 As shown, the top left is the metallic target to be imaged. Figure 7 The top right image shows the two-dimensional focused imaging result obtained at the distance plane where the target is located. Figure 7 Below are the results of three-dimensional volume imaging or volume data projection based on multiple distance slices.

[0061] exist Figure 7In the two-dimensional imaging result shown in the upper right corner, the horizontal axis "Horizontal" represents the target's spatial position in the horizontal direction, corresponding to the mechanical scanning direction of the millimeter-wave radar; the vertical axis "Vertical" represents the target's spatial position in the vertical direction, corresponding to the equivalent aperture direction of the receiving array. The unit for both the horizontal and vertical axes is millimeters (mm). This two-dimensional image is an imaging slice at a selected distance, that is, the horizontal-vertical reflection intensity distribution obtained after phase compensation and coherent focusing are completed near the target depth.

[0062] Two-dimensional images use grayscale to represent the millimeter-wave scattering intensity at different locations on the target. Brighter grayscale values ​​indicate a larger normalized echo amplitude at that location, typically corresponding to the surface, edges, transitions, and strong scattering areas facing the radar of metallic targets. Darker grayscale values ​​indicate a smaller echo amplitude at that location, corresponding to background areas, weak scattering areas, or areas where no effective target echo was detected. Therefore, the brighter areas in the image outline the main contours and key structural features of the metallic target.

[0063] Figure 7 The 3D volume imaging results below use pseudo-color and grayscale to represent the echo intensity at different spatial locations. Red and orange areas represent strong reflected echoes, gray areas represent weaker reflected echoes, and white or near-background colors represent low echo amplitudes or those below the display threshold. The three spatial dimensions of the 3D volume data correspond to the horizontal, vertical, and range directions, respectively, with the range direction representing the target's depth relative to the radar. By stacking and intensity-filtering 2D imaging slices from different range units, the distribution of the target's scattering centers in 3D space can be obtained.

[0064] Figure 7 The corresponding imaging scheme is as follows: based on the actual non-uniform scanning position fed back or recorded by the mechanical scanning platform, irregular spatial sampling coordinates are constructed; the original radar ADC data is sequentially rearranged, range-direction processed, channel amplitude and phase calibrated, multi-static to monostatic phase compensation is performed, non-uniform aperture density weighting is applied, and near-field wavenumber domain phase compensation is performed; then, non-uniform fast Fourier transform (NUFFT) is used to map the non-uniform spatial sampling data to the regular spatial frequency domain to complete the reconstruction of two-dimensional and three-dimensional near-field SAR images.

[0065] In summary, the above solutions can directly utilize the non-uniformly spaced sampling data generated by the mechanical scanning platform, reducing imaging defocus and structural distortion caused by scanning speed fluctuations, position errors, or uneven trajectories; at the same time, they can improve the focusing degree of strong scattering points of the target, enhance the recognizability of the main outline, edges, and local structures of the metal target, and realize the two-dimensional outline restoration and three-dimensional spatial positioning of irregularly shaped metal targets.

[0066] Furthermore, prior to the non-uniform fast Fourier transform mapping, the method further includes: assigning a weight value to each sampling point based on the distribution density of each sampling point in its neighborhood, and using the weight value to perform amplitude weighting processing on the echo data.

[0067] According to some embodiments of this disclosure, for each sampling point in a non-uniform aperture sampling point set, the distribution of sampling points in its surrounding neighborhood is statistically analyzed. If the sampling density of a sampling point in a certain region is low, a larger weight value is assigned to it; if the sampling density of a sampling point in a certain region is high, a smaller weight value is assigned to it. By introducing aperture density weighting or sampling weight compensation, the impact of non-uniform sampling density differences on frequency domain mapping and imaging stability can be improved.

[0068] In some other embodiments of this disclosure, when the sampling uniformity is different in different aperture directions, a fast Fourier transform can be used in the uniform sampling direction and a non-uniform fast Fourier transform can be used in the non-uniform sampling direction to complete the spatial spectrum transformation in a hybrid manner.

[0069] In other embodiments of this disclosure, when the mechanical scanning in the horizontal direction meets the uniform sampling condition, a Fast Fourier Transform (FFT) is used in the horizontal direction; when the virtual array in the vertical direction does not meet the uniform sampling condition, a Non-Uniform FFT is used in the vertical direction. When non-uniform sampling exists in both the horizontal and vertical directions, a two-dimensional non-uniform FFT is used for processing. This hybrid imaging method can further reduce the computational load and improve imaging efficiency while maintaining the adaptability of the non-uniform array.

[0070] The above solution will be further explained below with reference to the embodiments.

[0071] Figure 8 and Figure 9 The two-dimensional and three-dimensional imaging results obtained using the RMA method are shown below. Figure 10 and Figure 11 The two-dimensional and three-dimensional imaging results obtained using the NUFFT method are shown below. Figure 12 The left side shows an actual experimental scenario of mechanical scanning and target occlusion. Figure 12 The right side shows the actual scene imaging results obtained using the scheme proposed in this disclosure. In the two-dimensional imaging results, the horizontal axis "Horizontal" represents the horizontal position in the direction of radar mechanical scanning, and the vertical axis "Vertical" represents the vertical position in the direction of the array's equivalent aperture. The unit of coordinates is millimeters (mm). In the three-dimensional imaging results, the X-axis represents the horizontal position, the Y-axis represents the vertical position, and the Z / Depth axis represents the distance or depth position of the target relative to the radar. The three coordinate axes together describe the position of the target scattering point in three-dimensional space.

[0072] Figures 8 to 12The colors in the graph represent the normalized echo amplitude, with the unit of color scale being decibels (dB), and the displayed dynamic range is approximately [value missing]. 25–0 dB. Red corresponds to the strongest scattering region close to 0 dB, orange and yellow to relatively strong scattering regions, green and cyan to medium or weak scattering regions, and blue to weak scattering or background regions. Echoes below the lower limit of the display dynamic range are truncated or uniformly displayed as dark blue. This color is only used to indicate relative scattering intensity and does not represent the true color of the target surface.

[0073] A comparison of the above figures reveals that when using the conventional RMA method to directly process irregular scan data, the assumption of equal-interval aperture sampling positions leads to a mismatch between the actual non-uniform sampling data and the regular spatial frequency transformation conditions. This results in significant diffusion, sidelobes, and artifacts in the imaging results, making it difficult to accurately distinguish the edges and internal structures of the letter targets. Furthermore, the scattering points in the 3D results exhibit broadening and spatial ambiguity. However, using the NUFFT imaging method proposed in this paper, the overall outline, edges, and letter intervals of the letter "BIT" are clearly restored, and stray responses outside the target are significantly reduced. The 3D imaging results show a more concentrated target scattering distribution at the corresponding depth. This result demonstrates that NUFFT can perform spatial frequency transformation based on the actual non-uniform scan coordinates, reducing the impact of non-uniform sampling on coherent accumulation and phase focusing.

[0074] Figure 12 The experiment further illustrates a real-world imaging experiment under obstructed conditions. The millimeter-wave radar uses a mechanical scanning platform to scan targets inside cardboard boxes or other non-metallic obstructions using irregular trajectories. Figure 12 In the pseudo-color imaging results on the right, the red and yellow areas correspond to strongly reflective targets or metallic scattering parts in the scene, while the cyan and blue areas correspond to occlusions, background, and weakly scattering areas. This allows for the identification of the approximate outline of strongly scattering targets under occlusion conditions, and the determination of their position in the lateral, longitudinal, and distance directions.

[0075] The working principle of the technical solution corresponding to the above figures is as follows: the actual scanning position is used to replace the ideal equidistant aperture position, and non-uniform spatial sampling data transformation is performed by combining aperture density compensation and NUFFT; on this basis, range processing, channel calibration, multi-base phase compensation and near-field wavenumber domain focusing are performed to finally obtain two-dimensional range slices and three-dimensional volume imaging results.

[0076] Compared with directly using regular FFT or RMA to process irregular scan data, this solution has at least the following advantages: it can adapt to changes in scanning platform speed, positional jitter, and non-uniform sampling without forcing the original data to be treated as regular sampling data; it reduces interpolation errors and phase errors caused by non-uniform sampling, and reduces side lobes, ghosting, and false scattering points; it improves the focus of the target contour and the spatial positioning accuracy, making the edges of letters and internal intervals clearer; it supports two-dimensional slice imaging and three-dimensional volume imaging, and can simultaneously obtain the horizontal, vertical, and depth information of the target; and it utilizes the penetrating power of millimeter waves through non-metallic materials such as cardboard boxes to achieve the detection and positioning of strongly scattering targets under obstruction conditions.

[0077] Based on the same technological concept, such as Figure 13 As shown, this disclosure also proposes a millimeter-wave MIMO radar imaging system based on non-uniform sampling, comprising: The coordinate determination module is used to determine the coordinates of the transmitting antenna, the receiving antenna, and the virtual array elements; The coordinate acquisition module is used to acquire the coordinates of the mechanical scanning position in the global imaging coordinate system; The set construction module is used to superimpose the coordinates of virtual array elements with the coordinates of mechanical scanning positions to construct a set of non-uniform aperture sampling points; The data mapping module is used to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point using a non-uniform fast Fourier transform. The data compensation module is used for near-field phase compensation of echo data in the wavenumber domain; The image reconstruction module is used to reconstruct the radar reflection intensity image of the target area from the compensated echo data in the wavenumber domain.

[0078] This disclosure also proposes a millimeter-wave MIMO radar, including a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the above-described millimeter-wave MIMO radar imaging method based on non-uniform sampling.

[0079] This disclosure also proposes a computer program product stored in a computer-readable storage medium, which, when executed by a processor, is used to at least implement the above-described millimeter-wave MIMO radar imaging method based on non-uniform sampling.

[0080] Although the present disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A millimeter-wave MIMO radar imaging method based on non-uniform sampling, characterized in that, include: Determine the coordinates of the transmitting antenna, the receiving antenna, and the virtual array elements; Obtain the coordinates of the mechanical scanning position in the global imaging coordinate system; By superimposing the coordinates of the virtual array elements with the coordinates of the mechanical scanning position, a set of non-uniform aperture sampling points is constructed. A non-uniform fast Fourier transform is used to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point. Near-field phase compensation of echo data in the wavenumber domain; The radar reflection intensity image of the target area is reconstructed from the compensated echo data in the wavenumber domain.

2. The method according to claim 1, characterized in that, The coordinates of the virtual array elements are determined by the coordinates of the transmitting antenna and the receiving antenna according to the phase center, and all virtual array element positions are arranged at non-equal intervals; the mechanical scanning position coordinates are determined according to at least one of the following methods: slide rail motion parameters, trigger timing, or external encoder measurement values.

3. The method according to claim 1, characterized in that, The non-uniform aperture sampling point set is composed of MIMO channels and scanning positions; the non-uniformity of the non-uniform aperture sampling point set comes from at least one of the following factors: non-uniform virtual array, mechanical scanning step size error, non-equal interval step size, sparse sampling strategy, missing channels, and irregular array element positions.

4. The method according to claim 1, characterized in that, Before using the non-uniform fast Fourier transform for mapping, the following is also included: Equivalent monostatic phase compensation is performed on the MIMO multistatic channel to convert the multistatic propagation distance into the monostatic propagation distance at the virtual array element; the compensation factor of the equivalent monostatic phase compensation is constructed based on the phase difference between the multistatic propagation distance and the monostatic propagation distance at the virtual array element.

5. The method according to claim 1, characterized in that, Using a non-uniform fast Fourier transform, the echo data on the set of non-uniform aperture sampling points is mapped to a regular wavenumber domain grid based on the true spatial coordinates of each sampling point, including: For each range wavenumber sampling point, the spatial spectrum on the regular wavenumber domain grid is calculated based on the real spatial coordinates of each sampling point to achieve the mapping of echo data.

6. The method according to claim 1, characterized in that, The near-field phase compensation is based on the spherical wave propagation model, and the wavenumber domain data is phase-corrected according to the wavenumber in the depth direction. The reconstructed radar reflection intensity image of the target area includes: The focusing results at wavenumber sampling points at different distances are coherently accumulated, and the reflection intensity image of the target area at a specified depth is obtained by two-dimensional inverse Fourier transform.

7. The method according to claim 1, characterized in that, Prior to the non-uniform fast Fourier transform mapping, the method further includes: assigning a weight value to each sampling point based on the distribution density of each sampling point in its neighborhood, and using the weight value to perform amplitude weighting processing on the echo data.

8. A millimeter-wave MIMO radar imaging system based on non-uniform sampling, characterized in that, include: The coordinate determination module is used to determine the coordinates of the transmitting antenna, the receiving antenna, and the virtual array elements; The coordinate acquisition module is used to acquire the coordinates of the mechanical scanning position in the global imaging coordinate system; The set construction module is used to superimpose the coordinates of virtual array elements with the coordinates of mechanical scanning positions to construct a set of non-uniform aperture sampling points; The data mapping module is used to map the echo data on the set of non-uniform aperture sampling points to a regular wavenumber domain grid based on the real spatial coordinates of each sampling point using a non-uniform fast Fourier transform. The data compensation module is used for near-field phase compensation of echo data in the wavenumber domain; The image reconstruction module is used to reconstruct the radar reflection intensity image of the target area from the compensated echo data in the wavenumber domain.

9. A millimeter-wave MIMO radar, characterized in that, It includes a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the millimeter-wave MIMO radar imaging method based on non-uniform sampling as described in any one of claims 1-7.

10. A computer program product, said computer program product being stored in a computer-readable storage medium, characterized in that, When the computer program product is executed by a processor, it is used to implement at least the millimeter-wave MIMO radar imaging method based on non-uniform sampling as described in any one of claims 1-7.

Images