Radar three-dimensional imaging method and device based on area array angular domain decomposition

Abstract

The invention provides a radar three-dimensional imaging method and device based on area array angular domain decomposition, and the method comprises the steps: collecting original two-dimensional area array data, and carrying out the transformation processing, and obtaining first data; performing angular domain decomposition on the first data, dividing an overall view field into more than two local angular domain sub-regions, constructing a corresponding local angular domain decomposition weight function for each local angular domain sub-region, and performing weighting processing on the first data by using the local angular domain decomposition weight functions to obtain a weighted first data; obtaining second data of each local angular domain sub-region; performing local wavenumber domain three-dimensional imaging processing on each piece of second data to obtain a three-dimensional sub-image of a corresponding local angular domain sub-region; and fusing the three-dimensional sub-images of all the local angular domain sub-regions to obtain a final three-dimensional imaging result. While the advantage of high efficiency of wavenumber domain processing is maintained, decomposition modeling is performed on propagation differences in different angular domain regions, and the three-dimensional imaging quality under near-field, large-view-field and complex geometric scenes is improved.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and in particular to a radar three-dimensional imaging method and apparatus based on area array angular domain decomposition. Background Technology

[0002] Vehicle-mounted millimeter-wave radar has been widely used in fields such as intelligent driving, parking assistance, roadside perception, and spatial environment reconstruction due to its advantages such as all-weather operation, strong resistance to rain and fog, low cost, and ease of integration. As perception tasks gradually evolve from traditional point target detection to target contour recovery, 3D obstacle localization, road boundary modeling, and spatial occupancy estimation, radar systems are placing higher demands on their 3D imaging capabilities. Especially in vehicle-mounted forward-looking and surround-view scenarios, targets are often located at medium to close range and have complex spatial distributions; relying solely on traditional range-azimuth 2D imaging is no longer sufficient to meet the requirements for high-precision environmental perception.

[0003] Among existing vehicle-mounted radar 3D imaging technologies, common methods include azimuth-elevation beamforming based on 2D arrays, 3D imaging based on back projection, and wavenumber domain reconstruction methods that draw inspiration from SAR / near-field scanning imaging. Wavenumber domain imaging methods, in particular, show promising engineering application prospects because they can transform the spatial propagation model into a frequency domain processing problem and achieve efficient imaging through FFT, interpolation, and inverse transform.

[0004] However, for vehicle-mounted millimeter-wave radar, especially systems employing two-dimensional area arrays or two-dimensional MIMO virtual arrays, three-dimensional imaging still faces the following objective difficulties:

[0005] First, a two-dimensional area array simultaneously carries spatial sampling information in both horizontal and vertical dimensions, and target echoes are coupled across azimuth, elevation, and range dimensions. If a unified wavenumber domain model is directly established on the entire two-dimensional aperture, model mismatch is likely to occur under large field of view, near-field, or large oblique-view conditions. Second, the limited aperture, constrained installation attitude, and large target range of vehicle-mounted radar arrays result in inconsistent propagation characteristics in different angular domain regions. Directly performing a unified three-dimensional wavenumber domain reconstruction across the entire aperture can easily lead to local focus degradation, sidelobe lift, and decreased spatial resolution. Third, if a full three-dimensional spatial voxel-level backprojection method is used to avoid model mismatch, the computational load increases significantly, making it difficult to meet the real-time processing requirements of vehicle-mounted systems.

[0006] To address the aforementioned difficulties, existing technologies generally suffer from the following shortcomings:

[0007] First, traditional wavenumber domain 3D imaging methods typically assume that the array response satisfies a uniform propagation model across the entire field of view, without considering model deviations in wide forward field of view and near-field scenarios.

[0008] Secondly, traditional two-dimensional array processing methods usually perform two-dimensional FFT or uniform interpolation on all array data, failing to correct or reconstruct the propagation differences in different angular regions.

[0009] Third, under near-field and large field of view conditions, relying solely on a single global wavenumber domain mapping relationship will lead to significant focusing errors and imaging distortions in the edge regions of the field of view.

[0010] Fourth, while some high-precision methods can improve imaging results, they usually require large-scale 3D search or voxel-by-voxel coherent accumulation, which has high computational complexity and is not conducive to real-time deployment on vehicles.

[0011] Therefore, there is an urgent need for a radar three-dimensional imaging method and device based on area array angular domain decomposition to improve the above problems. Summary of the Invention

[0012] The purpose of this invention is to provide a radar three-dimensional imaging method and device based on the decomposition of the area array corner domain. While maintaining the high efficiency advantage of wavenumber domain processing, it can decompose and model the propagation differences of the two-dimensional area array in different corner domain regions, thereby improving the three-dimensional imaging quality in near field, large field of view and complex geometric scenes.

[0013] In a first aspect, the present invention provides a radar three-dimensional imaging method based on area array angular domain decomposition, comprising the steps of: acquiring original two-dimensional area array data and performing transformation processing to obtain first data; performing angular domain decomposition on the first data to divide the overall field of view into two or more local angular domain sub-regions, and constructing a corresponding local angular domain decomposition weight function for each local angular domain sub-region; using the local angular domain decomposition weight function to perform weighted processing on the first data to obtain second data for each local angular domain sub-region; performing local wavenumber domain three-dimensional imaging processing on each second data to obtain a three-dimensional sub-image of the corresponding local angular domain sub-region; and fusing the three-dimensional sub-images of all local angular domain sub-regions to obtain the final three-dimensional imaging result.

[0014] Optionally, acquiring raw two-dimensional array data and performing transformation processing to obtain first data includes: acquiring raw two-dimensional array data and transforming the raw two-dimensional array data to the range frequency domain through fast time Fourier transform to obtain first data; and / or the raw two-dimensional array data is two-dimensional array data or two-dimensional virtual array data of vehicle radar.

[0015] Optionally, performing angular domain decomposition on the first data to divide the overall field of view into two or more local angular domain sub-regions includes: dividing the overall field of view into two or more local angular domain sub-regions based on the azimuth and elevation field of view ranges, each local angular domain sub-region having a corresponding center direction; and / or constructing a corresponding local angular domain decomposition weight function for each of the local angular domain sub-regions includes: for the first For each local corner sub-region, the local corner decomposition weight function is constructed as follows:

[0016]

[0017] in, It is a positive integer; Windowing function for array; The imaginary unit; The center wave number; For the first Unit vectors in the direction of the center of each local angular sub-region; The coordinates of the array element positions. Indicates the horizontal position of the array. Indicates the vertical position of the array.

[0018] Optionally, weighting the first data using the local angular domain decomposition weight function includes: multiplying the local angular domain decomposition weight function with the first data; and / or the fusion is coherent fusion, incoherent fusion, or adaptive weighted fusion, wherein the weights of the adaptive weighted fusion are set according to the local signal-to-noise ratio, focus sharpness, spectral energy distribution, or main lobe and side lobe indices.

[0019] Optionally, performing local wavenumber domain three-dimensional imaging processing on each second data point to obtain a three-dimensional sub-image of the corresponding local angular domain sub-region includes: performing a Fourier transform on each second data point in the two-dimensional array dimension to obtain a two-dimensional spatial spectrum within the corresponding local angular domain sub-region; mapping the two-dimensional spatial spectrum to the three-dimensional wavenumber domain according to the local wavenumber domain mapping relationship to obtain non-uniformly sampled three-dimensional wavenumber domain data; performing regularized resampling on the non-uniformly sampled three-dimensional wavenumber domain data to obtain three-dimensional wavenumber domain data on a regular grid; and performing a three-dimensional inverse Fourier transform on the three-dimensional wavenumber domain data on the regular grid to obtain a three-dimensional sub-image of the corresponding local angular domain sub-region.

[0020] Optionally, the local wavenumber domain mapping relationship is as follows:

[0021]

[0022] in, and For two-dimensional spatial frequency, For frequency The corresponding wave number, This represents the wave number in the depth direction.

[0023] Secondly, the present invention provides a radar three-dimensional imaging device based on area array angular domain decomposition, the device comprising modules / units for performing any of the possible design methods described in the first aspect above. These modules / units can be implemented in hardware or by hardware executing corresponding software.

[0024] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a program executable on the processor, and when the program is executed by the processor, the electronic device implements a method for performing any of the possible designs described above.

[0025] Fourthly, the present invention provides a readable storage medium storing a program, which, when executed, implements a method of any possible design of any of the above aspects.

[0026] Fifthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0027] The beneficial effects of the method of this invention are as follows: Based on the original two-dimensional array data, the observation space of the two-dimensional array is first decomposed into angular domains, dividing the overall field of view into two or more local angular domain sub-regions; then, corresponding local wavenumber domain imaging models are established in each local angular domain sub-region, and local wavenumber domain mapping, focusing, and three-dimensional image reconstruction are completed; finally, the imaging results of each local sub-region are fused to obtain the final three-dimensional imaging result. The two-dimensional array wavenumber domain imaging problem, which was originally processed uniformly in the entire field of view, is transformed into a wavenumber domain imaging sub-problem in multiple local angular domains. Since the range of each local angular domain is small, its propagation characteristics are closer to the stationary local model, thereby reducing model mismatch in vehicle near-field, large squint, and wide field of view scenarios, and improving three-dimensional focusing capability. Thus, while maintaining the high efficiency advantage of wavenumber domain processing, the propagation differences of the two-dimensional array in different angular domain regions are decomposed and modeled, thereby improving the three-dimensional imaging quality in near-field, large field of view, and complex geometric scenarios. Attached Figure Description

[0028] Figure 1 A flowchart illustrating a radar three-dimensional imaging method based on area array angular domain decomposition provided in an embodiment of the present invention;

[0029] Figure 2 A schematic diagram of the structure of a radar three-dimensional imaging device based on area array angular domain decomposition provided in an embodiment of the present invention;

[0030] Figure 3This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed following the word and its equivalents, but do not exclude other elements or objects.

[0032] The technical solutions of the embodiments of the present invention will be described below with reference to the accompanying drawings. In the description of the embodiments of the present invention, the terminology used in the following embodiments is for the purpose of describing specific embodiments only and is not intended to limit the present invention. The singular expressions “a,” “the,” “the,” and “this” are intended to also include expressions such as “one or more,” unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of the present invention, “at least one” and “one or more” refer to one or more (including two). The term “and / or” is used to describe the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character “ / ” generally indicates that the preceding and following related objects are in an “or” relationship.

[0033] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of the invention. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," and "in still other embodiments" appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless otherwise specifically emphasized. The term "connection" includes both direct and indirect connections, unless otherwise stated. "First" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated.

[0034] In embodiments of the present invention, "exemplarily" or "for example" are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design described as "exemplarily" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or design solutions. Rather, the use of "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.

[0035] like Figure 1 As shown, this invention provides a radar three-dimensional imaging method based on area array angular domain decomposition, including the following steps:

[0036] S101: Collect the original two-dimensional area array data and perform transformation processing to obtain the first data.

[0037] In some embodiments, to facilitate subsequent angular domain decomposition, acquiring original two-dimensional area array data and performing transformation processing to obtain first data includes: acquiring original two-dimensional area array data and transforming the original two-dimensional area array data to the range frequency domain through fast time Fourier transform to obtain first data.

[0038] In other embodiments, the original two-dimensional array data is the two-dimensional array data of the vehicle-mounted radar or the two-dimensional virtual array data.

[0039] S102, perform angular domain decomposition on the first data, divide the overall field of view into two or more local angular domain sub-regions, construct a corresponding local angular domain decomposition weight function for each local angular domain sub-region, and use the local angular domain decomposition weight function to perform weighted processing on the first data to obtain the second data of each local angular domain sub-region.

[0040] In some embodiments, performing angular domain decomposition on the first data to divide the overall field of view into two or more local angular domain sub-regions includes: dividing the overall field of view into two or more local angular domain sub-regions based on the azimuth field of view range and the elevation field of view range, with each local angular domain sub-region having a corresponding center direction. Through angular domain decomposition, echo data with similar propagation characteristics are restricted to processing within the local sub-regions, thereby making the propagation models within each sub-region more consistent and reducing the accumulation of model errors.

[0041] In other embodiments, constructing a corresponding local corner domain decomposition weight function for each of the said local corner domain sub-regions includes:

[0042] Regarding the first For each local corner sub-region, the local corner decomposition weight function is constructed as follows:

[0043]

[0044] in, It is a positive integer; Windowing function for array; The imaginary unit; The center wave number; For the first Unit vectors in the direction of the center of each local angular sub-region; The coordinates of the array element positions. Indicates the horizontal position of the array. The vertical position of the array is represented by the corner decomposition weight function constructed in each local corner sub-region. By performing weighting processing on the original two-dimensional array data, the main propagation phase corresponding to the center direction of the local corner is shifted or compensated to the vicinity of the baseband in advance, so that the remaining phase gradient in the local corner is reduced and the complex response change between array elements is smoother.

[0045] In some other embodiments, weighting the first data using the local corner domain decomposition weight function includes multiplying the local corner domain decomposition weight function with the first data.

[0046] In some embodiments, the fusion is coherent fusion, incoherent fusion, or adaptive weighted fusion, and the weights of the adaptive weighted fusion are set according to local signal-to-noise ratio, focus sharpness, spectral energy distribution, or main lobe and side lobe indices.

[0047] S103, perform local wavenumber domain three-dimensional imaging processing on each second data point to obtain the corresponding local angular domain three-dimensional sub-image.

[0048] In some embodiments, performing local wavenumber domain three-dimensional imaging processing on each second data point to obtain a three-dimensional sub-image of the corresponding local angular domain sub-region includes: performing a Fourier transform on each second data point in a two-dimensional array dimension to obtain a two-dimensional spatial spectrum within the corresponding local angular domain sub-region; mapping the two-dimensional spatial spectrum to a three-dimensional wavenumber domain according to a local wavenumber domain mapping relationship to obtain non-uniformly sampled three-dimensional wavenumber domain data; performing regularized resampling on the non-uniformly sampled three-dimensional wavenumber domain data to obtain three-dimensional wavenumber domain data on a regular grid; and performing a three-dimensional inverse Fourier transform on the three-dimensional wavenumber domain data on the regular grid to obtain a three-dimensional sub-image of the corresponding local angular domain sub-region.

[0049] S104, fuse the three-dimensional sub-images of all local corner regions to obtain the final three-dimensional imaging result.

[0050] In some embodiments, the local wavenumber domain mapping relationship is as follows:

[0051]

[0052] in, and For two-dimensional spatial frequency, For frequency The corresponding wave number, This refers to the wave number in the depth direction (i.e., the radar line-of-sight direction).

[0053] The advantages of this invention are as follows: Based on the original two-dimensional array data, the observation space of the two-dimensional array is first decomposed into angular domains, dividing the overall field of view into two or more local angular domain sub-regions; then, corresponding local wavenumber domain imaging models are established in each local angular domain sub-region, and local wavenumber domain mapping, focusing, and three-dimensional image reconstruction are completed; finally, the imaging results of each local sub-region are fused to obtain the final three-dimensional imaging result. The original two-dimensional array wavenumber domain imaging problem, which was uniformly processed across the entire field of view, is transformed into a wavenumber domain imaging sub-problem within multiple local angular domains. Since each local angular domain is relatively small, its propagation characteristics are closer to the stationary local model, thereby reducing model mismatch in near-field, large squint, and wide-field-of-view scenarios, and improving three-dimensional focusing capability. Thus, while maintaining the high efficiency advantage of wavenumber domain processing, the propagation differences of the two-dimensional array in different angular domain regions are decomposed and modeled, thereby improving the three-dimensional imaging quality in near-field, large-field-of-view, and complex geometric scenarios.

[0054] To facilitate understanding, this embodiment further elaborates on the specific implementation process of the above method in conjunction with a specific application scenario. Taking the two-dimensional area array data collected by vehicle-mounted radar as an example, the specific steps include:

[0055] I. Array and Signal Models

[0056] Let the positions of the elements of the two-dimensional surface array be:

[0057]

[0058] in, Indicates the horizontal position of the array. This indicates the vertical position of the array. The target spatial position is denoted as:

[0059]

[0060] Where: 𝑋 is the horizontal coordinate; 𝑌 is the forward distance coordinate; 𝑍 is the height coordinate.

[0061] The propagation distance from the target to the array element is:

[0062]

[0063] For a single-station FMCW radar, after demodulation and range compression, the received signal at the corresponding range frequency point can be written as:

[0064]

[0065] in: The target space scattering coefficient; This represents the single-pass wavenumber; because it's a single-station two-pass transmission, the phase term is... .

[0066] This equation shows that the complex signals measured by the two-dimensional array at different frequencies are essentially the integral superposition of the target's three-dimensional scattering function under the spherical wave propagation kernel.

[0067] II. Propagation distance unfolded at the local corner center

[0068] To facilitate wavenumber domain processing, in the local angular sub-region The center direction is selected as:

[0069]

[0070] The corresponding unit vector for the central observation direction is denoted as:

[0071]

[0072] Suppose that the targets within this local angular domain are mainly distributed at the center slant distance. Nearby, the transmission distance can be increased. Expand around the local center direction. Let the target's position relative to the reference center be offset as follows:

[0073]

[0074] Under the local small-angle approximation, the propagation distance can be expressed as:

[0075]

[0076] in, This represents a local higher-order residual term.

[0077] Substituting this into the expression for the received signal, we get:

[0078]

[0079] Separating the terms related to the array element positions and those related to the target positions, we obtain:

[0080]

[0081] Due to the local angular domain Because it is relatively small, this expression is closer to a stationary local wavenumber domain model than a uniform processing across the entire field of view.

[0082] III. Mathematical Expression of Corner Domain Decomposition

[0083] Introduce a local corner domain decomposition weight function to the overall two-dimensional area array data:

[0084]

[0085] in: Windowing function for array; The center wave number.

[0086] Multiplying the original data by this weighting function is equivalent to shifting the main propagation phase corresponding to the center direction of the local angular domain to the baseband, resulting in:

[0087]

[0088] The physical significance of this process lies in: for the direction of the center of the corner domain The nearby echoes have their main angular phase gradients canceled out, making the remaining phase in this local region change more slowly with the array elements, which facilitates subsequent two-dimensional spatial frequency transformation and local wavenumber domain reconstruction.

[0089] IV. Local Two-Dimensional Spatial Frequency Transformation

[0090] Data after local corner domain decomposition Performing a Fourier transform on the two-dimensional array dimension yields:

[0091]

[0092] In the form of a discrete array, the corresponding form is:

[0093]

[0094] at this time, It represents the two-dimensional spatial spectrum within a local angular subregion.

[0095] V. Local Wavenumber Domain Mapping Relationship

[0096] Since single-station two-way propagation satisfies the total wavenumber constraint:

[0097]

[0098] Therefore, we can conclude that:

[0099]

[0100] This means that for each frequency and each local two-dimensional spatial frequency point All of these can be mapped to a point in the three-dimensional K-space. .

[0101] Therefore, the data in the local corner sub-region can be obtained from

[0102]

[0103] Mapped to

[0104]

[0105] because It is a non-linear relationship, therefore It is generally three-dimensional wavenumber domain data with non-uniform sampling.

[0106] VI. Local K-space resampling and 3D reconstruction

[0107] Let the regular three-dimensional wavenumber grid be:

[0108]

[0109] Then, interpolation resampling is performed on locally non-uniformly sampled data:

[0110]

[0111] in, This is the interpolation kernel function.

[0112] Then, a three-dimensional inverse Fourier transform is performed on the regularized local K-space data:

[0113]

[0114] This will give you the three-dimensional image corresponding to the local angular domain.

[0115] 7. Multi-angle domain fusion

[0116] Finally, all local corner sub-images are fused:

[0117]

[0118] in, For weight fusion.

[0119] If incoherent fusion is used, it can be written as:

[0120]

[0121] If coherent fusion is used, it can be written as:

[0122]

[0123] Processing flow

[0124] Step 1: Acquire raw two-dimensional array data of the vehicle-mounted radar

[0125]

[0126] Step 2: Fast Time Fourier Transform (FFT) to obtain the distance-frequency domain data (i.e., the first data).

[0127]

[0128] Step 3: Accumulate slow time or select imaging frames

[0129]

[0130] Step 4: Divide the two-dimensional array into corner domains and construct local corner domain decomposition weight functions.

[0131] Step 5: Obtain the second data of the local corner sub-region.

[0132]

[0133] Step 6: Perform a two-dimensional spatial frequency transformation on each local angular domain.

[0134]

[0135] Step 7: Complete according to local wavenumber domain relationship Mapping

[0136] Step 8: Perform local K-space regularization resampling

[0137] Step 9: Perform a local 3D IFFT to obtain...

[0138] Step 10: Fuse the local images and output the final 3D imaging result.

[0139] The key to the embodiments of the present invention lies in:

[0140] (1) The problem of three-dimensional imaging of the entire field of view of a two-dimensional array is decomposed into multiple local angular domain wavenumber domain imaging sub-problems. The method of this invention does not directly establish a unified three-dimensional wavenumber domain imaging model for the entire field of view of the two-dimensional array, but divides the overall field of view into multiple local angular domain sub-regions according to the differences in the propagation characteristics of the two-dimensional array of the vehicle radar in the azimuth-elevation joint field of view, and establishes a corresponding local wavenumber domain imaging model for each local angular domain. Its technical essence is: to transform the problem that was originally handled by "one global large model" into a problem that is handled by "multiple local small models".

[0141] This process addresses the problem of propagation model mismatch that arises in traditional full-field-of-view unified modeling under wide field of view, near-field, and large oblique-view conditions in vehicular scenarios. Because the target propagation path, phase curvature, and spatial frequency distribution differ significantly in different field-of-view directions within a vehicular scene, using a single global model often results in good focusing in some areas while focusing degrades in others. The method of this invention, through angular domain decomposition, restricts echo data with similar propagation characteristics to local sub-regions for processing, thereby making the propagation model more consistent within each sub-region and reducing model error accumulation.

[0142] This structural design is not simply data partitioning, but is closely related to subsequent local wavenumber domain mapping, local K-space resampling, and local inverse transform imaging, forming a complete "decomposition-reconstruction-fusion" chain. This is the primary core element that distinguishes the method of this invention from traditional unified wavenumber domain imaging methods.

[0143] (2) The local main propagation phase shift is completed by using the corner domain decomposition weight function to achieve local phase stabilization. The method of the present invention constructs a corresponding corner domain decomposition weight function in each local corner domain sub-region. By performing weighting processing on the original two-dimensional array data, the main propagation phase corresponding to the center direction of the local corner domain is shifted or compensated to the vicinity of the baseband in advance, so that the remaining phase gradient in the local corner domain is reduced and the complex response change between array elements is smoother.

[0144] The key to this step lies not in simple windowing, but in introducing a phase weighting function that matches the local center direction, transforming the original data from a "wide-field signal with rapidly changing phase" into a "locally phase-gradiently changing signal." Mathematically, this is equivalent to using the local viewpoint center as a reference propagation direction to perform directional demodulation of the two-dimensional array response; physically, it is equivalent to first aligning the local viewpoint in the data domain before performing subsequent wavenumber domain reconstruction.

[0145] The direct effects of this process are twofold: firstly, it helps improve the spectral concentration after local two-dimensional spatial frequency transformation and reduces spectral spread; secondly, it helps reduce numerical errors during local K-space mapping and interpolation, thereby improving the focusing stability of local sub-images. Compared to traditional methods that directly perform unified two-dimensional FFT or unified K-space interpolation on the original full-field data, the method of this invention completes angular domain localization before entering the wavenumber domain processing, thus allowing for more relaxed error control conditions in subsequent steps.

[0146] (3) Establish a local wavenumber domain mapping relationship within the local angular domain to reduce global nonlinear mapping error. The method of this invention does not use a single [method / mechanism] within the overall two-dimensional field of view. Instead of a fixed mapping relationship, a corresponding local wavenumber domain mapping relationship is established for each local angular sub-region, taking into account its central propagation direction and local approximation conditions. Since the angular range covered by each local angular domain is limited and the target propagation direction is relatively concentrated, the relationship between the two-dimensional spatial frequency components and the depth wavenumber components within this local region is closer to a locally linear or locally weakly nonlinear form.

[0147] This process addresses a key issue with traditional global mapping in wide-field-of-view vehicular scenarios: the inconsistent adaptation of the same global mapping relationship to the center and edges of the field of view leads to significantly larger reconstruction errors in the edge regions compared to the center region. This is particularly problematic for near-field targets and targets with large oblique views; directly using a globally unified wavenumber domain relationship can easily result in inaccurate depth wavenumber estimation, thus affecting the 3D focusing position and main lobe convergence performance.

[0148] The method of this invention uses a "local angular domain corresponding to local mapping" approach to make the K-space sampling relationship more closely resemble the real propagation mechanism in each sub-region. This not only improves the imaging accuracy of local sub-images but also enhances the overall 3D reconstruction quality of edge fields of view and targets with complex poses.

[0149] (4) K-space regularization resampling is performed on each local angular domain to reduce interpolation distortion and spectral distortion. In wavenumber domain imaging, the original three-dimensional K-space samples are usually not distributed in a regular rectangular grid, so they need to be resampled to a regular grid before the three-dimensional inverse Fourier transform can be performed. Traditional methods often perform global K-space interpolation on all data under the overall full field of view. However, the method of this invention performs regularization resampling on the wavenumber domain data of each local angular domain after angular domain decomposition.

[0150] This is a crucial technical aspect of the invention. K-space interpolation error is related not only to the interpolation algorithm itself, but also to the curvature of the sampling distribution of the interpolated data, the local spectral spread range, and the rate of phase change. After angular domain decomposition, the local data has a more concentrated spectral distribution and a smoother curvature change in the sampling surface within the local range, making it more suitable for low-error interpolation. In other words, the method of this invention is not merely "block interpolation," but rather improves the interpolation conditions from the signal structure level by first reducing the geometric complexity of local samples through angular domain decomposition and then performing regularized resampling.

[0151] The benefits of this process are significant: it reduces errors caused by nonlinear mapping, finite support of the interpolation kernel, and spectral leakage during K-space resampling; it helps improve the fidelity of edge fields of view and weakly scattering targets; and it can also reduce local energy stretching and spectral distortion caused by a single global interpolation.

[0152] (5) The imaging framework of “local reconstruction-global fusion” is adopted to balance accuracy and complete field of view coverage. After the method of the present invention independently completes the three-dimensional wavenumber domain imaging in each local angular domain, it does not take a certain local sub-image as the final result alone. Instead, it introduces a multi-angular domain sub-image fusion mechanism to perform coherent fusion, incoherent fusion or adaptive weighted fusion on each local image to form the final three-dimensional imaging result.

[0153] The key to this design lies in the fact that each local angular sub-image has high focusing accuracy in its corresponding direction. However, individual local sub-images typically only respond optimally to targets within a specific local angular domain. Without fusion, it is difficult to cover the entire field of view or to accommodate targets in all directions. Through a fusion mechanism, this invention combines the local advantages of each local imaging sub-problem into an overall advantage, thereby ensuring full field of view coverage while inheriting the high accuracy of each local model.

[0154] Furthermore, the fusion weights can be adaptively set based on local signal-to-noise ratio, focus sharpness, spectral energy distribution, or main lobe and side lobe indices, thus enabling further suppression of anomalous sub-images and enhancement of high-quality sub-image contributions. This makes the present invention not merely a "locally processed stitching" but a multi-sub-image collaborative reconstruction framework with a globally optimal trend.

[0155] (6) Establish a wavenumber domain decomposition processing link suitable for vehicle-mounted scenarios, taking into account both near-field adaptability and real-time implementation. Traditional high-precision 3D imaging methods mostly rely on back projection or voxel-by-voxel spherical wave matching accumulation. Although these methods provide accurate modeling, the computational load increases rapidly with the number of image voxels, array elements, and frequency sampling points, which is not conducive to real-time implementation on vehicle-mounted platforms. The method of this invention introduces angular domain decomposition within the wavenumber domain framework, so that the overall processing link still maintains the regular structure of "two-dimensional array spectrum transformation - K-space resampling - three-dimensional IFFT".

[0156] The significance of this process lies in the fact that the method of this invention does not completely abandon the high efficiency advantage of the wavenumber domain in order to improve imaging accuracy. Instead, while retaining the fast implementation framework of FFT, it improves accuracy by adding two key steps: "corner domain decomposition" and "local reconstruction." Therefore, the core contribution of the method of this invention is not simply to pursue theoretical accuracy, nor simply to pursue engineering speed, but to establish a more balanced accuracy-complexity trade-off in the context of vehicle-mounted near-field 3D imaging.

[0157] This feature makes the method of the present invention easier to deploy on FPGA, DSP, GPU or heterogeneous SoC platforms, and facilitates modular design, parallel scheduling and pipelined processing in engineering implementation.

[0158] (7) The array structure characteristics, field-of-view decomposition strategy, and wavenumber domain reconstruction mechanism are designed collaboratively. The method of this invention does not use the two-dimensional area array merely as an ordinary spatial sampler, but rather combines the azimuth-elevation two-dimensional sampling characteristics of the area array structure itself to propose an angular domain decomposition strategy, and makes this decomposition strategy work together with subsequent local two-dimensional spatial frequency transformation, local wavenumber domain mapping, and local image fusion. In other words, the method of this invention is not an innovation at a single algorithm point, but a collaborative design of the entire chain from "array observation structure - local data organization - frequency domain reconstruction - result fusion".

[0159] This collaborative design approach has significant patent value because it demonstrates that the method of this invention is not simply a modification of existing wavenumber domain methods, but rather a systematic solution addressing the core conflict between "wide field of view and high precision" in the specific context of two-dimensional area array vehicle-mounted three-dimensional imaging.

[0160] This systematic processing chain makes the method of this invention more feasible and has clearer technical boundaries.

[0161] The advantages of the embodiments of the present invention are as follows:

[0162] I. Significantly improves focusing consistency in vehicle-mounted wide-field-of-view 3D imaging

[0163] Traditional global unified wavenumber domain imaging methods typically achieve good focusing in the central region of the field of view. However, as the target deviates from the main viewing direction, propagation model errors gradually accumulate, easily leading to main lobe widening, side lobe lifting, and spatial positioning deviations in the peripheral angular domains. This invention decomposes the field of view into angular domains and reconstructs the wavenumber domain separately in local angular domains, making the propagation model in each angular domain closer to real-world conditions. Therefore, it can significantly improve focusing consistency across different field of view regions.

[0164] Specifically, this invention maintains more balanced imaging clarity for targets at the center, edge, and different elevation regions of the field of view; avoids the problem of "clear center, blurry edge" in traditional methods; and makes the three-dimensional imaging results more stable, continuous, and reliable throughout the entire working field of view. This effect is crucial for vehicle-mounted radar because targets in real-world road environments are not always directly in front of the array but may be widely distributed across the entire field of view.

[0165] II. Improve the 3D positioning accuracy of near-field and mid-near-field targets

[0166] Key obstacles commonly encountered by vehicle-mounted millimeter-wave radars, such as vehicles ahead, guardrails, curbs, pedestrians, and irregularly shaped obstacles at close range, are mostly located in the near-field or mid-near-field region. In these areas, spherical wave characteristics are pronounced, and using a globally uniform far-field or weakly near-field approximation model can easily lead to spatial focusing errors. This invention reduces local approximation errors through local angular domain modeling, resulting in more accurate positioning of near-range targets in the lateral, vertical, and depth directions.

[0167] The corresponding benefits include: reduced drift of the target scattering center in the 3D image; target contour recovery that more closely approximates the true geometry; and enhanced spatial separation capability between multiple targets. For road environment modeling and obstacle boundary extraction, this means that subsequent detection, clustering, and tracking modules can obtain more reliable 3D input.

[0168] III. Reduce K-space interpolation errors under wide field-of-view conditions and enhance image fidelity.

[0169] In wavenumber domain imaging, K-space regularization interpolation is often a significant step in introducing errors. Traditional global interpolation schemes are prone to energy diffusion, amplitude distortion, and phase deviation in the interpolated spectral data due to the large sample distribution span and rapid changes in local curvature. This invention addresses this by employing a "first angular domain decomposition, then local regularization" approach, making the data within each local sub-region more suitable for interpolation, thereby reducing resampling errors.

[0170] This benefit manifests in the final image as follows: the main lobe of the target is more concentrated, and the side lobes are lower; weak scattering centers are less likely to be overwhelmed by global interpolation errors; and image details are better preserved under different local viewpoints. Especially when it is necessary to recover slender targets, edge structures, or structures with abrupt changes in local height, this invention can preserve more spatial details than traditional uniform interpolation methods.

[0171] IV. Approaching higher precision imaging results without employing voxel-by-voxel back projection.

[0172] In high-precision 3D imaging, back projection methods are often considered to produce high-quality images, but at the cost of extremely high computational complexity. This invention employs a local wavenumber domain reconstruction approach after angular domain decomposition. While maintaining the fast FFT-type processing framework, it introduces local propagation adaptation capabilities, thus achieving a focusing effect closer to high-precision imaging without resorting to large-scale voxel-by-voxel back projection.

[0173] The corresponding benefits are: while achieving significantly lower computational complexity and processing latency compared to the back projection method, it still obtains good 3D focusing clarity and spatial positioning accuracy. This makes the present invention more practically valuable in automotive embedded real-time systems, rather than merely remaining at the level of offline high-precision experiments.

[0174] V. Improve the imaging robustness of complex scenes and targets from non-central perspectives.

[0175] In real-world road environments, targets often exhibit complex scattering characteristics, such as vehicle sides, angled parked vehicles, irregularly shaped roadside obstacles, metal guardrails, and steps. The scattering responses of these targets can vary significantly across different angular domains. Traditional global unified models often respond well in some directions but poorly in others, leading to unstable overall imaging results. This invention, by performing local processing on different angular domains and then coordinating reconstruction through subsequent fusion, can better adapt to the imaging needs of multi-directional scattering centers in complex scenes.

[0176] Its benefits are reflected in: more robust focusing on yaw targets, pitch-changing targets, and targets at the edge of the viewpoint; the ability to rely on other local sub-images to provide supplementary information even when the quality of some local sub-images deteriorates; and a higher tolerance of the overall system to model mismatch, local signal-to-noise ratio fluctuations, and local angular domain scattering inhomogeneities.

[0177] VI. Significantly improves the real-time performance and feasibility of project deployment.

[0178] This invention is still based on the wavenumber domain fast imaging framework. The main computations consist of local two-dimensional spatial frequency transformation, local K-space resampling, and local three-dimensional IFFT, which are suitable for parallel and pipelined implementation. Compared with voxel-by-voxel back projection, its computational structure is more regular; compared with globally unified high-precision modeling, it can be processed in blocks; and compared with a single full-field-of-view model, it is easier to flexibly configure the number of local angular domains and the processing granularity according to the hardware computing power.

[0179] Therefore, this invention offers the following direct benefits in engineering implementation: First, it facilitates parallel deployment on platforms such as FPGA, GPU, and DSP; second, the number of angular domain partitions can be selected based on system computing power to achieve adjustable "precision-complexity"; and third, it is more suitable for automotive scenarios requiring high frame rates, low power consumption, and stable latency. In other words, this invention not only improves imaging quality but also enhances the feasibility of practical deployment.

[0180] like Figure 2As shown, based on the above method, the present invention provides a radar three-dimensional imaging device based on area array angular domain decomposition, comprising: an acquisition unit 201, used to acquire original two-dimensional area array data and perform transformation processing to obtain first data; a decomposition unit 202, used to perform angular domain decomposition on the first data, dividing the overall field of view into multiple local angular domain sub-regions, and constructing a corresponding local angular domain decomposition weight function for each local angular domain sub-region, and using the local angular domain decomposition weight function to perform weighted processing on the first data to obtain second data for each local angular domain sub-region; a processing unit 203, used to perform local wavenumber domain three-dimensional imaging processing on each second data to obtain a three-dimensional sub-image of the corresponding local angular domain sub-region; and a fusion unit 204, used to fuse the three-dimensional sub-images of all local angular domain sub-regions to obtain the final three-dimensional imaging result.

[0181] It should be understood that all relevant content of each step involved in the above method embodiments can be referenced to the functional description of the corresponding functional module, and will not be repeated here. Furthermore, the use of suffixes such as "module," "component," or "unit" to represent elements is merely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, "module," "component," or "unit" can be used interchangeably. Terminals can be implemented in various forms. For example, the terminals described in this invention may include mobile terminals such as mobile phones, tablets, laptops, handheld computers, personal digital assistants (PDAs), portable media players (PMPs), navigation devices, wearable devices, smart bracelets, pedometers, etc., as well as fixed terminals such as digital TVs and desktop computers. The following description will use mobile terminals as examples; those skilled in the art will understand that, in addition to elements specifically designed for mobile purposes, the construction according to embodiments of the present invention can also be applied to fixed-type terminals.

[0182] In other embodiments of the present invention, an electronic device 300 is disclosed, such as... Figure 3 As shown, the device may include: one or more processors 301; memory 302; display 303; one or more application programs (not shown); and one or more computer programs 304. These devices can be connected via one or more communication buses 305. The one or more computer programs 304 are stored in the memory 302 and configured to be executed by the one or more processors 301. The one or more computer programs 304 include instructions that can be used to perform actions such as... Figure 1 Each step in the corresponding embodiment.

[0183] Processor 301 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0184] The memory 302 can be an internal storage unit of the electronic device 300, such as a hard disk or RAM of the electronic device 300. The memory 302 can also be an external storage device of the electronic device 300, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or FlashCard equipped on the electronic device 300. Furthermore, the memory 302 can include both internal and external storage units of the electronic device 300. The memory 302 is used to store computer programs and other programs and data required by the electronic device. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0185] The computer program 304 can be divided into one or more modules / units. The one or more modules / units can be a series of computer program instruction segments that can perform a specific function. The instruction segments are used to describe the execution process of the computer program 304 in the electronic device 300.

[0186] In addition to the above-described structure, those skilled in the art will understand that Figure 3 This is merely an example of electronic device 300 and does not constitute a limitation on electronic device 300. Electronic device 300 may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0187] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the functions described above can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0188] Based on the above embodiments, the present invention also discloses a computer-readable storage medium having at least one computer program stored thereon, wherein the computer program, when executed by a processor, implements the methods described in the foregoing embodiments.

[0189] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a computer-readable storage medium, which is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof. The storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. This available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state drive (SSD)).

[0190] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0191] Although the embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. The above descriptions are merely embodiments of the present invention and do not limit the patent scope of the present invention. However, it should be understood that such modifications and variations fall within the scope and spirit of the present invention. Moreover, the present invention described herein may have other embodiments and can be implemented or realized in various ways. All equivalent transformations made based on the description and drawings of the present invention, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A radar three-dimensional imaging method based on area array angular domain decomposition, characterized in that, Including the following steps: The original two-dimensional area array data is collected and transformed to obtain the first data. The first data is decomposed into a angular domain, dividing the overall field of view into two or more local angular domain sub-regions. A corresponding local angular domain decomposition weight function is constructed for each local angular domain sub-region. The first data is weighted using the local angular domain decomposition weight function to obtain the second data for each local angular domain sub-region. Each second data point is processed using local wavenumber domain three-dimensional imaging to obtain a corresponding three-dimensional sub-image of the local angular domain sub-region. The three-dimensional sub-images of all local corner regions are fused to obtain the final three-dimensional imaging result.

2. The method according to claim 1, characterized in that, The original two-dimensional area array data is acquired and transformed to obtain the first data, which includes: The original two-dimensional array data is acquired, and the original two-dimensional array data is transformed to the range frequency domain through fast time Fourier transform to obtain the first data; And / or the original two-dimensional array data is the two-dimensional array data or two-dimensional virtual array data of the vehicle radar.

3. The method according to claim 1 or 2, characterized in that, Angular domain decomposition is performed on the first data, dividing the overall field of view into two or more local angular domain sub-regions, including: Based on the azimuth and elevation field of view, the overall field of view is divided into two or more local angular sub-regions, each with a corresponding central direction. And / or constructing a corresponding local corner domain decomposition weight function for each of the said local corner domain sub-regions includes: Regarding the first For each local corner sub-region, the local corner decomposition weight function is constructed as follows: in, It is a positive integer; Windowing function for array; The imaginary unit; The center wave number; For the first Unit vectors in the direction of the center of each local angular sub-region; The coordinates of the array element positions, Indicates the horizontal position of the array. Indicates the vertical position of the array.

4. The method according to claim 1, characterized in that, The weighting process of the first data using the local corner domain decomposition weight function includes: The local corner domain decomposition weight function is multiplied by the first data; And / or the fusion is coherent fusion, incoherent fusion or adaptive weighted fusion, and the weights of the adaptive weighted fusion are set according to the local signal-to-noise ratio, focus sharpness, spectral energy distribution or main lobe and side lobe indices.

5. The method according to claim 1, characterized in that, Each second data point is processed using local wavenumber domain 3D imaging to obtain a corresponding 3D sub-image of the local angular domain sub-region, including: Perform a Fourier transform on each second data point in the two-dimensional array dimension to obtain the two-dimensional spatial spectrum within the corresponding local angular sub-region; Based on the local wavenumber domain mapping relationship, the two-dimensional spatial spectrum is mapped to the three-dimensional wavenumber domain to obtain non-uniformly sampled three-dimensional wavenumber domain data; The non-uniformly sampled three-dimensional wavenumber domain data is resampled in a regularized manner to obtain three-dimensional wavenumber domain data on a regular grid. A three-dimensional inverse Fourier transform is performed on the three-dimensional wavenumber domain data on the regular grid to obtain the three-dimensional sub-image of the corresponding local angular domain sub-region.

6. The method according to claim 5, characterized in that, The local wavenumber domain mapping relationship is as follows: in, and For two-dimensional spatial frequency, For frequency The corresponding wave number, This represents the wave number in the depth direction.

7. A radar three-dimensional imaging device based on area array angular domain decomposition, used in the method according to any one of claims 1-6, characterized in that, include: The acquisition unit is used to acquire the original two-dimensional area array data and perform transformation processing to obtain the first data; The decomposition unit is used to perform angular domain decomposition on the first data, divide the overall field of view into multiple local angular domain sub-regions, construct a corresponding local angular domain decomposition weight function for each local angular domain sub-region, and use the local angular domain decomposition weight function to perform weighted processing on the first data to obtain the second data of each local angular domain sub-region. The processing unit is used to perform local wavenumber domain three-dimensional imaging processing on each second data to obtain a three-dimensional sub-image of the corresponding local angular domain sub-region. The fusion unit is used to fuse the three-dimensional sub-images of all local corner regions to obtain the final three-dimensional imaging result.

8. An electronic device, characterized in that, The device includes a memory and a processor, wherein the memory stores a program that can run on the processor, and when the program is executed by the processor, causes the electronic device to perform the method of any one of claims 1-6.

9. A readable storage medium storing a program, characterized in that, When the program is executed, it implements the method of any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-6.

Images