A method, device, storage medium and program product for relocating non-line-of-sight building layouts in three-dimensional imaging of an array interferometric synthetic aperture radar
An initial 3D point cloud is generated by an array interferometric synthetic aperture radar system. Combined with the Alpha Shape algorithm and density clustering, the layout of non-line-of-sight buildings is identified and corrected, solving the problem of non-line-of-sight building relocation in urban environments and achieving high-precision 3D reconstruction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2025-05-28
- Publication Date
- 2026-05-19
AI Technical Summary
Existing technologies cannot effectively identify and correct non-line-of-sight building layouts in urban environments, resulting in multipath ghosting in 3D imaging and reducing the reliability of imaging results. In particular, they cannot achieve relocation of non-line-of-sight buildings when the building layout is unknown.
Multi-view observations were conducted using an array interferometric synthetic aperture radar system to generate an initial 3D point cloud. The surface was reconstructed using the Alpha Shape algorithm. Combined with density clustering and plane fitting algorithms, reflection points of non-line-of-sight points were identified and corrected. Mirror symmetry geometry was used to reposition the building layout.
It can accurately identify and correct non-line-of-sight building layouts without prior structural information, improve the accuracy of 3D imaging, enhance the accuracy of urban perception and structural reconstruction, and is highly adaptable to complex urban environments.
Smart Images
Figure CN120630201B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of array interferometric synthetic aperture radar three-dimensional imaging technology, and specifically relates to a non-line-of-sight building layout imaging technology. Background Technology
[0002] Array interferometric synthetic aperture radar (AAP) 3D imaging technology primarily utilizes radar to observe urban targets from different angles. Combined with multi-baseline interferometric synthesis processing, it achieves high-precision 3D reconstruction of buildings and terrain in complex urban environments, possessing significant application value in urban planning, building structure monitoring, and post-disaster assessment. When radar signals propagate in an urban environment, they reach the target and return to the receiver along multiple paths. The signal returning to the receiver after directly illuminating the target is called the direct path signal. Due to the dense high-rise buildings and smooth walls in cities, radar electromagnetic waves undergo one or more reflections on building exterior surfaces after emission, forming multipath signals that return to the receiver after reflection from walls or the ground. These signals may also reflect off other building surfaces after passing the target and then bounce back to the receiver, ultimately causing multipath ghosting in the imaging. Multipath ghosting severely interferes with the quality of 3D point clouds, reducing the reliability of imaging results. Through proper identification and correction of multipath ghosting, target reconstruction can be achieved even in areas lacking direct paths, thereby enhancing the completeness and accuracy of urban 3D modeling. Therefore, the utilization of multipath signals is crucial in urban environmental perception under 3D imaging conditions.
[0003] In the field of 3D imaging, research on multipath exploitation methods for urban environmental perception is still limited and in its early stages. The Aerospace Information Research Institute of the Chinese Academy of Sciences proposed a non-line-of-sight target imaging method based on multipath signal analysis (Y.Lin, X.Qiu, Y.Luo, et al., “Viability of multipath exploitation in urban canyon: Range profiles analysis in TomoSAR imaging,” IGARSS, 2024.). This method demonstrates the feasibility of using multipath signals for non-line-of-sight target imaging by analyzing the propagation characteristics of multipath signals in non-line-of-sight regions. By analyzing the number of reflections and paths of multipath signals, it can detect and image non-line-of-sight targets hidden in urban canyons. Building upon this foundation, the Aerospace Information Research Institute of the Chinese Academy of Sciences further proposed a method for relocating non-line-of-sight targets in 3D imaging of building areas using Synthetic Aperture Radar (SAR) based on a multipath model (Lin Yuqing, Qiu Xiaolan, Peng Lingxiao, et al. Relocating Non-line-of-sight Targets in 3D Imaging of Building Areas Using SAR Based on Multipath Model [J]. Journal of Radar. 2024.). This method utilizes the geometric relationships and reflection characteristics of multipath signals, establishes a low-altitude spherical wave model and a multipath model, and combines building plane fitting and geometric analysis to restore the virtual image of non-line-of-sight targets to their true position, thereby achieving accurate 3D imaging and relocation of non-line-of-sight targets in urban canyons. However, the above method can only handle scenarios where the building layout is known. When the building layout parameters are unknown, it cannot achieve non-line-of-sight building layout relocation. Therefore, researching methods for relocating non-line-of-sight building layouts in 3D imaging using array interferometric synthetic aperture radar has significant value in the field of urban environmental perception. Summary of the Invention
[0004] To address the aforementioned technical problems, this invention proposes a method, device, storage medium, and program product for relocating non-line-of-sight building layouts in three-dimensional imaging using array interferometric synthetic aperture radar. This method requires no prior information about the building structure and can identify and correct non-line-of-sight building ghost images caused by multipath propagation in urban environments.
[0005] One of the technical solutions adopted in this invention is: a method for relocating non-line-of-sight building layouts in three-dimensional imaging using array interferometric synthetic aperture radar, comprising:
[0006] S1. The target area is observed from multiple perspectives using an array interferometric synthetic aperture radar system. The tomographic synthetic aperture radar method is used to coherently combine multiple SAR images obtained from multiple observation angles, extract the scattering distribution of each pixel in the vertical direction, and generate the initial three-dimensional point cloud of the observation scene.
[0007] S2. Randomly downsample the initial 3D point cloud obtained in step S1, and then use the Alpha Shape algorithm to reconstruct the surface of the downsampled result to generate a 3D surface model that reflects the outer contour of the building structure.
[0008] S3. By analyzing whether the line connecting each point in the initial three-dimensional point cloud to the radar antenna intersects with the three-dimensional surface model generated in step S2, the propagation path type is distinguished, and the line-of-sight point, non-line-of-sight point, and the reflection point associated with the non-line-of-sight point are obtained.
[0009] S4. Correct the non-line-of-sight points obtained in step S3;
[0010] S5. Overlay the line-of-sight points obtained in step S3 with the corrected non-line-of-sight points obtained in step S4 to obtain the final 3D point cloud.
[0011] In step S1, an array interferometric synthetic aperture radar (SAR) system is used to conduct multi-view observations of the target area. A tomographic SAR method is employed to coherently combine multiple SAR images acquired from different observation angles, extracting the vertical scattering distribution of each pixel to generate an initial 3D point cloud of the observed scene. This 3D point cloud reflects the approximate spatial distribution of scatterers and can describe the spatial characteristics of building structures within the line-of-sight area. However, due to multipath propagation effects, the layout of buildings outside the line-of-sight area often appears in the point cloud as offsets, distortions, or ghosting, failing to reflect their true spatial location. Therefore, subsequent steps are required for geometric repositioning to achieve accurate reconstruction of the outside-line-of-sight building structures.
[0012] The sampling operation in step S2 can be described as follows:
[0013] P down ={p i |p i ∈P,ξ i <η}
[0014] Where P is the initial point cloud set, P down p is the set of downsampled point clouds. i Let ξ be a point in P. i Let η be a uniform random variable in the interval [0,1], and let η be the sampling rate.
[0015] The surface reconstruction in step S2 specifically involves using the Alpha Shape algorithm to reconstruct the surface from the downsampling results, generating a three-dimensional surface model that reflects the outer contour of the building structure, denoted as S. r This provides spatial boundary information for subsequent reflection path analysis and non-line-of-sight target relocation.
[0016] Step S3 mainly involves identifying non-line-of-sight building layouts and associated reflection points. The specific process is as follows:
[0017] The 3D point cloud obtained in Step 1 contains both line-of-sight (LOS) and non-line-of-sight (NOS) points. To identify NOS building layouts, the point cloud needs to be classified, and the reflection point associated with each NOS point needs to be determined. Since multipath propagation mainly involves reflections on surfaces (this method only considers first-order reflections), the propagation path type can be distinguished by analyzing whether the line connecting each point to the radar antenna intersects with the surface model.
[0018] Let the position of the radar antenna be... Let x, y, z represent the set of real numbers, where x, y, z represent the three-dimensional spatial geometric coordinates of the antenna. Consider the following two cases, selecting any point P in the initial three-dimensional point cloud:
[0019] Case 1 (line segment path): If line segment AP intersects with surface S r If there is no intersection, then point P is the line-of-sight point.
[0020] Case 2 (Non-line-of-sight path): If line segment AP intersects with surface S r The existence of an intersection point R indicates that the radar signal reaches point P after being reflected from a point. Therefore, point P is the line-of-sight point, and point R is its associated reflection point. The position of R can be determined by the following expression: R∈S r And A+t(P) NLOS -A)=R, where t∈(0,1).
[0021] By traversing the entire point cloud and performing the above judgment on each point, all non-line-of-sight points and the associated set of reflection points can be identified.
[0022] The non-line-of-sight building layout correction process described in step S4 is as follows:
[0023] S41. Use a density clustering algorithm (such as DBSCAN) to cluster the reflection points identified in step S3 to form several reflection point clusters.
[0024] The mathematical expression of the DBSCAN algorithm is as follows:
[0025] set up Define a point as the set of reflection points. The neighborhood is:
[0026]
[0027] Where d(·,·) is the Euclidean distance, and ∈ is the minimum neighborhood radius. The conditions for being the core point are:
[0028]
[0029] Where |·| represents the cardinality of the set, and MinPts is the preset threshold for the minimum number of neighborhood points.
[0030] S42. Apply the RANSAC algorithm to each cluster for plane fitting, extract the corresponding reflection plane, and the fitted plane equation can be expressed as:
[0031] π k :a k x+b k y+c k z+d k =0
[0032] Among them, (a k ,b k ,c k ) is the plane normal vector And it satisfies the normalization condition:
[0033]
[0034] ‖·‖2 represents the L2 norm;
[0035] S43. Using the fitted plane as the axis of mirror symmetry, project the corresponding non-line-of-sight points to their mirror positions. For any non-line-of-sight point... and the reflecting plane π k Its mirror point The calculation can be expressed as follows:
[0036]
[0037] Where D is to plane π k Directed distance:
[0038]
[0039] By performing the above mirror transformation on all non-line-of-sight points, spatial correction and repositioning of non-line-of-sight building layouts can be achieved, thereby restoring their true geometric positions without relying on prior structural information.
[0040] The second technical solution adopted in this invention is: a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of a non-line-of-sight building layout relocation method in three-dimensional imaging of array interferometric synthetic aperture radar.
[0041] The third technical solution adopted in this invention is: a computer scale storage medium storing a computer program, which, when executed by a processor, implements the steps of a non-line-of-sight building layout relocation method in three-dimensional imaging of array interferometric synthetic aperture radar.
[0042] The third technical solution adopted in this invention is: a computer program product, including a computer program, which, when executed by a processor, implements the steps of a non-line-of-sight building layout relocation method in three-dimensional imaging of array interferometric synthetic aperture radar.
[0043] The beneficial effects of this invention are as follows: This invention proposes a method for relocating non-line-of-sight building layouts in 3D imaging using array interferometric synthetic aperture radar (ASAR). First, an initial 3D point cloud of the observation scene is acquired using an ASAR system. Then, the Alpha Shape algorithm is applied to reconstruct the surface of the point cloud, constructing a boundary model for reflection path analysis. Next, by analyzing whether the line connecting the point cloud and the radar antenna intersects with the surface model, non-line-of-sight points and their associated reflection points are identified. Finally, density clustering and plane fitting algorithms, combined with mirror symmetry geometry, are used to correct and relocate the non-line-of-sight building layout.
[0044] This invention provides a method that requires no prior structural information, effectively identifies multipath reflection paths, recovers the true spatial location of buildings in non-line-of-sight areas, and significantly improves 3D imaging accuracy, providing strong support for SAR applications such as urban sensing and structural reconstruction. This method can accurately identify and correct non-line-of-sight building ghost images generated by multipath propagation without any prior building structural information. By constructing a 3D point cloud and combining it with reflection point analysis, this method can effectively distinguish between line-of-sight points and non-line-of-sight points, and utilizes mirror symmetry geometry to reposition offset non-line-of-sight building structures to their true locations. Compared to existing multipath correction methods that rely on prior geometric models, this invention has stronger adaptability and can obtain high-precision, high-reliability 3D reconstruction results in complex urban environments, providing strong support for SAR systems in practical applications such as urban sensing and building monitoring. Attached Figure Description
[0045] Figure 1 Flowchart for non-line-of-sight building layout relocation method.
[0046] Figure 2 This is a schematic diagram of a simulation scenario.
[0047] Figure 3 This is a schematic diagram of the antenna location distribution.
[0048] Figure 4 This is the initial 3D point cloud side view.
[0049] Figure 5 This is the initial top view of the 3D point cloud.
[0050] Figure 6 It is a three-dimensional curved surface model.
[0051] Figure 7 It is a set of non-line-of-sight points.
[0052] Figure 8 This is the set of reflection points.
[0053] Figure 9 This is the fitting result for the reflective surface.
[0054] Figure 10 The 3D point cloud after repositioning. Detailed Implementation
[0055] To facilitate understanding of the technical content of this invention by those skilled in the art, the following description, in conjunction with the accompanying drawings, further illustrates the invention.
[0056] This embodiment uses, as follows Figure 2 Using the dimensions and CAD model of the L-shaped building shown as an example, the starting coordinates of the UAV-borne radar's trajectory are (-28m, -300m, 300m), and the ending coordinates are (88m, -300m, 300m). The azimuth sampling interval is 0.08m, and the radar's incident angle is 45°. The radar's transmitted signal is a stepped-frequency signal with a center frequency of 16GHz and a bandwidth of 1600MHz, with a frequency step of 0.8MHz. The radar array mode is 1 transmit and 16 receive, and the antenna positions are distributed as follows... Figure 3 As shown in the figure. In this simulation scenario, radar echo data is generated using an electromagnetic scattering modeling method based on SBR-PO.
[0057] like Figure 1 As shown, the method of the present invention includes the following processing steps:
[0058] Step 1: Acquiring 3D Point Clouds
[0059] The 1-transmit, 16-receive mode is equivalent to 16 antenna phase centers, enabling the acquisition of 16 single-look SAR images. After preprocessing the images, such as registration and de-skewing, an orthogonal matching pursuit algorithm is used for sparse reconstruction in the height direction to obtain an initial 3D point cloud. The reconstruction result is shown below. Figure 4 and Figure 5 As shown. Figure 5 In the diagram, the area circled in dashes behind the L-shaped building is the building's NLOS surface, or non-visual distance surface. Its position needs to be corrected to achieve repositioning of the non-visual distance building layout.
[0060] Step 2: Surface Reconstruction
[0061] To reduce the computational complexity of subsequent steps, the initial 3D point cloud obtained in step 1 is randomly downsampled according to a set ratio. The sampling ratio must be selected to balance point cloud density and algorithm efficiency. An excessively large sampling ratio will increase the number of triangular facets and increase the computational burden, while an excessively small ratio will result in a sparse point cloud, affecting the quality of surface reconstruction. In this embodiment, the sampling ratio is set to 0.1. Subsequently, the Alpha Shape algorithm is used to reconstruct the surface of the downsampled point cloud. The Alpha sphere radius is set to 3, and the hole filling threshold is set to 1. The surface reconstruction result is as follows: Figure 6 As shown.
[0062] Step 3: Identification of non-line-of-sight building layout and associated reflection points
[0063] The point cloud dataset is traversed, and the propagation path type is determined based on whether the line connecting each point to the radar antenna intersects the surface model. If the line does not intersect the surface, the point is a line-of-sight point; if an intersection exists, it is a non-line-of-sight point, and the intersection is the associated reflection point of that point. The final non-line-of-sight point set and reflection point set are as follows: Figure 7 and Figure 8 As shown.
[0064] Step 4: Non-line-of-sight building layout correction
[0065] Density-Based Spatial Clustering of Applications with Noise (DBSCAN) was used to cluster the reflection points identified in step 3 and remove outliers. The neighborhood radius was set to 5, and the minimum number of neighborhood points (MinPts) was set to 10, resulting in only one cluster of reflection points. Subsequently, the Random Sample Consensus (RANSAC) algorithm was applied to this cluster for plane fitting, and the fitting result is shown below. Figure 9 As shown, the unnormalized equation of the reflection plane is -0.0216x - 16.3012y + 581.2563 = 0. Finally, by making the non-line-of-sight point set mirror-symmetric about the reflection plane, the repositioning of the non-line-of-sight building layout can be achieved. The repositioning result is as follows: Figure 10 As shown, the repositioning error is controlled within 0.5m.
[0066] Simulation results show that the non-line-of-sight building layout relocation method for three-dimensional imaging of array interferometric synthetic aperture radar provided by this invention can not only effectively utilize multipath but also does not rely on prior layout information, thus verifying the correctness and effectiveness of this invention.
[0067] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of a non-line-of-sight building layout relocation method in three-dimensional imaging of array interferometric synthetic aperture radar.
[0068] This embodiment also provides a computer scale storage medium storing a computer program, which, when executed by a processor, implements the steps of a non-line-of-sight building layout relocation method in three-dimensional imaging of array interferometric synthetic aperture radar.
[0069] This embodiment also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of a non-line-of-sight building layout relocation method in three-dimensional imaging with array interferometric synthetic aperture radar.
[0070] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of the claims of the invention.
Claims
1. A method for relocating non-line-of-sight building layouts in three-dimensional imaging using array interferometric synthetic aperture radar, characterized in that, include: S1. The target area is observed from multiple perspectives using an array interferometric synthetic aperture radar system. The tomographic synthetic aperture radar method is used to coherently combine multiple SAR images obtained from multiple observation angles, extract the scattering distribution of each pixel in the vertical direction, and generate the initial three-dimensional point cloud of the observation scene. S2. Randomly downsample the initial 3D point cloud obtained in step S1, and then use the Alpha Shape algorithm to reconstruct the surface of the downsampled result to generate a 3D surface model that reflects the outer contour of the building structure. S3. By analyzing whether the line connecting each point in the initial three-dimensional point cloud to the radar antenna intersects with the three-dimensional surface model generated in step S2, the propagation path type is distinguished, and the line-of-sight point, non-line-of-sight point, and the reflection point associated with the non-line-of-sight point are obtained. S4. Correct the non-line-of-sight points based on the reflection points associated with the non-line-of-sight points; Step S4 specifically includes the following sub-steps: S41. Use a clustering algorithm to cluster the reflection points obtained in step S3 to form several clusters of reflection points; S42. Perform plane fitting on each cluster and extract the corresponding reflection plane; S43. Using the fitting plane as the mirror axis of symmetry, project the corresponding non-line-of-sight points to the mirror position to obtain the corrected non-line-of-sight points. S5. Overlay the line-of-sight points obtained in step S3 with the corrected non-line-of-sight points obtained in step S4 to obtain the final 3D point cloud.
2. The method for non-line-of-sight building layout relocation in three-dimensional imaging using array interferometric synthetic aperture radar according to claim 1, characterized in that, The specific process of step S3 is as follows: Let the position of the radar antenna be... ; Select any point in the initial 3D point cloud Consider the following two cases: Case 1: If line segment With noodles If there is no intersection, then the point The point of view; The three-dimensional surface model obtained in step S2; Case 2: If line segment With noodles There are intersections This indicates that the radar signal reaches the point after being reflected from the point. Then point For non-line-of-sight points, point For its associated reflection point, The position is determined by the following expression: and ; in ; By traversing the entire initial 3D point cloud, all sets of line-of-sight points, sets of non-line-of-sight points, and sets of reflection points associated with non-line-of-sight points are identified.
3. The method for relocating non-line-of-sight building layouts in three-dimensional imaging using array interferometric synthetic aperture radar according to claim 2, characterized in that, Step S41 specifically employs a density clustering algorithm.
4. The method for non-line-of-sight building layout relocation in three-dimensional imaging using array interferometric synthetic aperture radar according to claim 3, characterized in that, Step S42 specifically uses the RANSAC algorithm for plane fitting.
5. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-4.
6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-4.
7. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-4.