Image generation system and method based on SAR satellite

Abstract

The invention discloses a multi-view image generation system and method based on an SAR satellite. The system comprises a data acquisition module, a preprocessing module, a data fusion module, an interference phase acquisition module, a three-dimensional reconstruction correction module, an error correction module and an image generation module. The system firstly obtains multi-view SAR satellite images of the same target area, and obtains high-resolution preprocessed images after slant range correction, motion compensation and focusing imaging. Accurate registration of the multi-view images is achieved through a cross-correlation algorithm, fusion is carried out through a weighted average or principal component analysis method, and fusion data containing multi-angle scattering features are obtained; and extracting an interference phase based on the fusion data, and performing inversion according to an interference measurement formula to obtain a target height. The corrected three-dimensional model is used for generating a final visual image of the target area. According to the scheme, building edge dislocation can be effectively improved, and imaging precision and model stability are improved.

Description

Technical Field

[0001] This invention relates to the field of satellite detection technology, and in particular to an image generation system and method based on SAR satellites. Background Technology

[0002] Synthetic Aperture Radar (SAR) is an active microwave remote sensing technology capable of all-weather, all-time imaging under complex weather conditions. It transmits pulse signals via a satellite platform and receives the scattered echoes, obtaining high-resolution images through pulse compression and coherent processing. Compared to optical imaging, SAR has irreplaceable advantages in military reconnaissance, topographic mapping, and disaster monitoring. However, in densely populated urban areas with tall buildings, SAR imaging encounters severe geometric distortion problems, especially overlay. Because echoes from the top of buildings arrive at the receiver before echoes from the bottom, the tops of buildings in the image appear displaced towards closer distances, sometimes even appearing as if the top and bottom are inverted. This distortion not only interferes with the accurate identification of building shapes but also severely affects the accuracy of retrieving three-dimensional parameters such as building height.

[0003] To address this issue, existing technologies have proposed methods such as stereo SAR, interferometric SAR (InSAR), and multi-baseline interferometric SAR. Stereo SAR relies on matching and reconstructing multi-view images, but in high-density urban areas, the complexity of scatterers makes matching extremely difficult. InSAR uses phase difference to invert altitude, performing well in mountainous areas, but it is prone to failure in areas with tall buildings due to multipath scattering. While multi-baseline InSAR can improve accuracy, it requires a sophisticated satellite formation system, resulting in high cost and complexity. In summary, existing technologies still struggle to effectively suppress the overlay effect in urban areas, thus necessitating a new method to achieve more accurate 3D reconstruction and image generation. Summary of the Invention

[0004] In view of the problems existing in the prior art, one aspect of the present invention is to provide a multi-view image generation system based on SAR satellites, characterized in that it includes:

[0005] The data acquisition module is used to acquire multi-view SAR satellite images of the same target area;

[0006] The preprocessing module, following the data acquisition module, is used to perform slant range correction, motion compensation, and focused imaging on the acquired multi-view SAR satellite images to generate high-resolution preprocessed images.

[0007] The data fusion module is used to register and fuse the preprocessed images to generate a comprehensive dataset containing multi-angle scattering features.

[0008] The interferometric phase acquisition module is used to extract interferometric phase information from the fused image;

[0009] The 3D reconstruction and correction module is used to generate 3D height information of the target region based on the interferometric phase.

[0010] An error correction module is used to correct and adjust the three-dimensional height information to ensure the accuracy of the three-dimensional reconstruction.

[0011] The image generation module is used to generate a final visualization image of the target area based on the three-dimensional height information.

[0012] Preferably, the slant range correction is used to reproject the slant distance of the fused image into a horizontal distance; the motion compensation is used to compensate for displacement errors caused by satellite or platform motion; and the focused imaging is used to generate a high-resolution target image.

[0013] Preferably, the data fusion module achieves accurate registration of multi-view images through a cross-correlation algorithm, and then uses a weighted average or principal component analysis method to fuse the registered images.

[0014] Preferably, the error correction module compensates the reconstructed 3D model. This module establishes a functional relationship between the residual error and the noise variance σ, the incident angle θ, and the baseline length B based on the formula: Δh=k1σ+k2tanθ+k3 / B. It fits the compensation coefficients k1, k2, and k3 using the least squares method and performs iterative optimization in combination with actual observation data.

[0015] Preferably, the data fusion module achieves accurate registration of multi-view images through a cross-correlation algorithm, ensuring that the error is controlled within 0.1 pixels.

[0016] Another aspect of the present invention is to provide a multi-view image generation method based on SAR satellites, comprising the following steps:

[0017] Acquire multi-view SAR images of the same target area;

[0018] The acquired images are subjected to slant range correction, motion compensation, and focusing imaging to generate high-resolution preprocessed images;

[0019] The preprocessed images are registered and fused to generate a comprehensive dataset containing multi-angle scattering features;

[0020] Extracting interferometric phase information from fused data;

[0021] Three-dimensional reconstruction is performed using height inversion formulas;

[0022] Residual error compensation is performed on the 3D height information obtained from the 3D reconstruction to obtain a corrected 3D height model;

[0023] The final target area visualization image is generated based on the three-dimensional height information obtained from the modified three-dimensional height model.

[0024] Preferably, the registration and fusion of the preprocessed images includes registration using a cross-correlation algorithm, followed by fusion using a weighted average or principal component analysis method.

[0025] Preferably, residual error compensation for the three-dimensional height model is performed based on the following formula:

[0026] Δh=k1σ+k2tanθ+k3 / B

[0027] Where Δh is the residual error, σ is the noise variance, θ is the incident angle, and B is the baseline length; the compensation coefficients k1, k2, and k3 are fitted by the least squares method, and the fitting results are iteratively optimized using actual observation data to obtain the final error-corrected three-dimensional model.

[0028] Preferably, the obtained corrected fused data includes the following steps: performing cross-correlation algorithm processing on multi-view images to achieve accurate registration; controlling the registration error within 0.1 pixels; and fusing the registered images using weighted average or principal component analysis methods to generate a comprehensive dataset.

[0029] Another aspect of the present invention is to provide a computer system for implementing any of the methods described above, including a memory, a processor, and a display device connected thereto, wherein the memory stores computer-executable instructions that, when executed by the processor, cause the system to perform any of the methods described above.

[0030] Compared with existing technical solutions, the present invention has at least the following beneficial effects:

[0031] 1) This invention acquires multiple images containing redundant geometric information by controlling SAR satellites to image the same area at multiple incident angles. Based on the fused multi-view SAR images, this invention uses interferometry to extract phase differences and combines them with a geometric correction model for height inversion. Building height information is obtained through height formula inversion, and an error model is used to compensate for the effects of multipath scattering and atmospheric delay, generating a high-precision 3D point cloud. Furthermore, the reconstructed 3D model is combined with the original SAR images to generate a distortion-corrected SAR image, thus ensuring accurate reconstruction of buildings in the images.

[0032] 2) Slant range correction aims to eliminate projection deviations caused by Earth's curvature and imaging geometry; motion compensation corrects phase errors caused by satellite orbital disturbances, ensuring consistency across multiple imaging sessions; imaging focusing employs the RD algorithm or Chirp Scaling algorithm to project the echoes onto a range-azimuth two-dimensional plane, yielding high-resolution complex SAR images. The processing results at this stage are crucial for ensuring that images from different angles can be compared and analyzed within a unified coordinate system. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of SAR satellite imaging provided by the present invention;

[0034] Figure 2 This invention provides an image generation system based on SAR satellites.

[0035] Figure 3 This is a flowchart of the SAR satellite image generation method provided by the present invention.

[0036] The present invention will now be described in further detail. However, the examples described below are merely simplified examples of the present invention and do not represent or limit the scope of protection of the present invention. The scope of protection of the present invention is determined by the claims. Specific Implementation

[0037] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0038] like Figure 2 The image shown is an image generation system based on SAR satellite provided by the present invention; the system includes: a data acquisition module, a preprocessing module, a data fusion module, an interferometric phase acquisition module, a three-dimensional reconstruction and correction module, an error correction module, and a final image generation module.

[0039] The data acquisition module controls the satellite to perform multiple imaging operations on the target area at different incident angles. For example... Figure 1 The diagram shown illustrates the SAR satellite imaging provided by this invention. By adjusting the orbital attitude and radar beam direction, multi-view echo signals of the same area can be acquired during multiple passes. These echo signals retain the scattering information of the target under different geometric projections, forming the basis for subsequent processing. After pulse compression, the signal's range resolution can reach the order of 1 meter, providing conditions for high-precision reconstruction.

[0040] The preprocessing module standardizes the raw echo (the raw echo signal acquired by the satellite in the data acquisition module of step S1). This process includes three sub-steps: slant range correction, motion compensation, and focused imaging.

[0041] 1) Slant range correction is used to reproject the slant distance of the original echo onto the horizontal distance coordinates to establish a unified ground geometric reference.

[0042] 2) Motion compensation is performed after slant range correction to correct real-time track deviations and attitude disturbances of the platform, so that the echo phase is consistent with the unified geometric coordinates;

[0043] 3) Focused imaging is performed after motion compensation. The RD algorithm or Chirp Scaling algorithm is used to perform joint compression of the range and azimuth directions to generate high-resolution complex SAR images.

[0044] Specifically, slant range correction is used to reproject the "slant range" from the curved Earth model onto the "horizontal distance" coordinate axis. Because satellite imaging uses an oblique observation method, different pixels correspond to different ground feature heights, resulting in a systematic offset between the slant range measurement position and the actual geographical location. By constructing an ellipsoidal model and combining imaging geometric parameters (incident angle, orbital parameters, azimuth position), this invention converts the slant range measurement of each pixel into a surface projection position, achieving preliminary geometric correction. This step can significantly suppress the forward shift effect of building tops caused by oblique imaging. It eliminates geometric projection errors in SAR oblique imaging mode, ensuring consistency between the echo measurement position and the actual surface position. Because SAR uses an oblique observation method to acquire echo signals, the distance recorded in the echo is the "slant range," which does not correspond to a horizontal ground projection. Due to the Earth's curvature, the horizontal surface distance corresponding to the same slant range is not consistent, leading to a systematic forward shift or compression in the uncorrected image.

[0045] To this end, this invention, based on an Earth ellipsoid model, utilizes satellite orbital parameters, incident angles, and imaging geometry to reproject the diagonal distance of each pixel onto the horizontal distance coordinate axis (ground range). Through this transformation, a true spatial relationship between pixels can be established, achieving a unified geometric benchmark.

[0046] Preferably, for each pixel (r) s ,θ s Slope distance measurement, ground projection point (x) g ,y g ) can be accessed

[0047] x g =Rsinθ s ,,y g =Rcosθ s get.

[0048] Motion compensation is used to correct phase errors caused by satellite orbital disturbances, ensuring consistency across multiple imaging sessions. During imaging, satellites are affected by factors such as attitude jitter, orbital errors, and atmospheric disturbances, causing random phase drift in the echo signal, manifested as blurring in the range and azimuth directions. This invention uses inertial navigation system (IMU) data, star sensor data, and high-precision orbital data (POD) to jointly estimate the satellite's instantaneous position, reprojecting the echo signal onto a unified reference orbital coordinate system. Motion compensation not only improves image focusing quality but, more importantly, ensures that multiple images acquired at different times and azimuth angles maintain a consistent phase reference during coherent processing, thereby improving the stability of interferometry from the source.

[0049] For each distance-direction sampling point, the phase φ c (r) Imaging focusing is used to focus the pulse-compressed echo signal into a high-resolution complex SAR image. Depending on the imaging mode, this invention employs either the Range-Doppler (RD) algorithm or the Chirp Scaling (CS) algorithm: the RD algorithm compensates for the azimuth Doppler center frequency and frequency modulation frequency to focus the echo in the azimuth direction; the CS algorithm achieves unified compensation for range-azimuth coupling errors by scaling the range pulse proportionally, making it particularly suitable for large oblique angle imaging scenarios. The RD or Chirp Scaling algorithm used in imaging focusing projects the echo onto a range-azimuth two-dimensional plane to obtain a high-resolution complex SAR image. The processing results at this stage are crucial for ensuring that images from different angles can be compared and analyzed in a unified coordinate system.

[0050] The data fusion module achieves accurate registration of multi-view images (obtained through the data acquisition module 10 mentioned above) through a cross-correlation algorithm, ensuring that the error is controlled within 0.1 pixels.

[0051] In a preferred embodiment, an example of using the cross-correlation algorithm to perform multi-view image registration includes: extracting grayscale values ​​or feature maps (such as edge or texture features) for each image to be registered; selecting a sliding window on the reference image and scanning the image to be registered with the same window; and calculating the correlation coefficient for each window position.

[0052] Then, weighted averaging or principal component analysis is used to fuse the registered images, resulting in a comprehensive dataset containing multi-angle scattering features. The advantage of this approach is that information that is occluded or distorted in a single viewpoint can be supplemented from other viewpoints, thereby mitigating the problem of building edge misalignment.

[0053] Specifically, the data fusion module processes the multi-view images obtained by the data acquisition module to obtain a comprehensive dataset containing multi-angle scattering features. Preferred processing steps include:

[0054] Image registration: The data fusion module first performs precise registration of multi-view images. Registration is achieved through a cross-correlation algorithm, which calculates the similarity of corresponding regions in different images at the pixel or sub-pixel level to determine the translational offset between images and aligns each image to a unified reference coordinate system, ensuring that the registration error is controlled within 0.1 pixels.

[0055] Image Fusion: For the registered images, the data fusion module uses one or more of the following methods for fusion: 1) Weighted Average: For each pixel location, the pixel value of the corresponding multiple images is multiplied by a preset weight, and then a weighted average is calculated to obtain the fused pixel value. The weights can be set according to the imaging angle, signal-to-noise ratio, or image quality to enhance reliable information and weaken noise. 2) Principal Component Analysis (PCA): Each registered image is converted into a data matrix composed of vectors. Principal component analysis is performed on the matrix to extract the principal components and reconstruct the image, thereby obtaining the fused image, which retains the common features of images from different viewpoints and suppresses noise and redundant information.

[0056] Through the above processing, the fused images can supplement the information that is occluded or distorted in a single viewpoint, alleviate the problem of building edge misalignment, and obtain more complete multi-angle scattering feature data.

[0057] The interferometric phase acquisition module extracts height information from the fused dataset. This module selects two or more coherent images and calculates the interferometric phase difference. The basic relationship of the interferometric phase is as follows:

[0058] Δφ=4πhBcosθ / λR

[0059] Where h is the target height, B is the baseline length, θ is the incident angle, and R is the slant range. Through phase filtering and unwrapping, a continuous phase distribution can be obtained, thereby indirectly retrieving height information.

[0060] To ensure stability, the interferometric processing also needs to calculate the coherence coefficient γ to assess the correlation between images. If γ is less than 0.3, it indicates that there is a strong decoherence effect in the region, which needs to be removed or the multi-view averaging method should be used to suppress noise.

[0061] After obtaining the interferometric phase, the three-dimensional reconstruction and correction module uses the height inversion formula.

[0062] h=λRΔφ / 4πBcosθ

[0063] Phase information is converted into height estimation (i.e., generating three-dimensional surface information from the two-dimensional image information in the above steps). Combined with ground control points (GCPs) for calibration, absolute height accuracy can be effectively improved.

[0064] By employing surface reconstruction methods such as Delaunay triangulation, discrete elevation point clouds are transformed into continuous 3D models, fully depicting the geometric structure of buildings. The reconstructed point cloud density can reach tens of points per square meter, which is sufficient to meet the needs of urban modeling.

[0065] However, in practical applications, high-rise building areas often suffer from error sources such as multipath scattering, atmospheric delay, and system noise. Therefore, the error correction module compensates for the reconstruction results. This module is based on the formula...

[0066] Δh=k1σ+k2tanθ+k3 / B

[0067] A functional relationship was established between residual error and noise variance σ, incident angle θ, and baseline length B. The compensation coefficients k1, k2, and k3 were fitted using the least squares method, and iterative optimization was performed using actual observation data, which significantly reduced the error. Experiments show that the traditional single-view SAR method for estimating building height in a city's CBD area has an error of approximately ±12 to 15 meters, while the method of this invention reduces the error to within ±3 meters.

[0068] Finally, the image generation module combines the corrected 3D model with the original SAR image to generate a SAR image with effectively corrected geometric distortion. In this stage, point cloud data is projected onto the image plane for orthorectification, thus eliminating inversions and misalignments caused by overlay. Compared with traditional images, the newly generated SAR image shows clear building boundaries, accurate top-to-bottom positioning, and improved consistency with optical images by more than 20%. In summary, the method of this invention realizes a complete technical chain from multi-angle data acquisition to final distortion-corrected image output. The modules are closely interconnected: data acquisition provides input for preprocessing, the preprocessed image is supplemented with redundant information in the fusion module, the fusion result enters the interferometry module to extract phase difference, height inversion generates a 3D point cloud, error correction improves accuracy, and finally the image generation module outputs the corrected SAR image. The entire process is interconnected, ensuring the acquisition of high-precision 3D structural information and imaging results even in complex urban environments.

[0069] Example 2

[0070] like Figure 3 The diagram shows a flowchart of the SAR satellite image generation method provided by the present invention, which includes the following steps:

[0071] Step S1: Control the satellite to perform multiple imaging operations on the target area at different incident angles;

[0072] Step S2: Standardize the raw echo of the satellite's transmitted beam to obtain a high-resolution complex SAR image;

[0073] Step S3: Achieve accurate registration of multi-view images through cross-correlation algorithm, and then fuse the registered images using weighted average or principal component analysis to obtain a comprehensive dataset containing multi-angle scattering features;

[0074] Step S4: Select two or more coherent images from the SAR imagery, calculate the interferometric phase difference, and use the height inversion formula to convert the phase information into a height estimate to obtain the reconstructed three-dimensional model.

[0075] Step S5: Combine the reconstructed 3D model with the SAR image to generate a SAR image with effectively corrected geometric distortion.

[0076] Through the above two innovations, the present invention has achieved remarkable results in experiments: In the experimental area of ​​a city center, the building height error obtained by the traditional single-view SAR method is about ±12 to 15 meters, while after adopting the method of the present invention, the height error is reduced to within ±3 meters; the phenomenon of "building tops moving forward and bottoms being obscured" that is common in the original SAR images is significantly alleviated, and the geometric structure of buildings in the generated images is more complete.

[0077] Compared with optical images, the SAR images generated by this invention show a more than 20% improvement in the consistency of building boundaries and clearer boundary contours. The method of this invention is well adapted to complex urban environments and can maintain stable imaging accuracy even in areas with high building density and large differences in building height, providing reliable data support for applications such as urban modeling and disaster monitoring.

[0078] The present invention also provides a computer system for implementing the above method, including a memory, a processor and a display device connected thereto. The memory stores computer-executable instructions, which, when executed by the processor, cause the system to perform the above steps, including data acquisition, preprocessing, fusion, error correction, interference phase extraction, three-dimensional reconstruction and image generation.

[0079] In summary, this invention takes "multi-angle observation and data fusion" and "three-dimensional reconstruction by interferometry combined with geometric correction" as its core innovations, solves the overlay problem that is common in high-rise building areas in existing technologies, and achieves the generation of SAR images with higher accuracy and more realistic geometric structure.

[0080] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0081] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

[0082] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. A multi-view image generation system based on SAR satellites, characterized in that, include: The data acquisition module is used to acquire multi-view SAR satellite images of the same target area; The preprocessing module, following the data acquisition module, is used to perform slant range correction, motion compensation, and focused imaging on the acquired multi-view SAR satellite images to generate high-resolution preprocessed images. The data fusion module is used to register and fuse the preprocessed images to generate a comprehensive dataset containing multi-angle scattering features. An interference phase acquisition module is used to extract interference phase information from the comprehensive dataset; A three-dimensional reconstruction and correction module is used to generate three-dimensional height information of the target region based on the interference phase. An error correction module is used to correct and adjust the three-dimensional height information to ensure the accuracy of the three-dimensional reconstruction. The image generation module is used to generate a final visualization image of the target area based on the corrected and adjusted three-dimensional height information.

2. The system according to claim 1, wherein, The slant range correction is used to reproject the slant distance of the fused image into a horizontal distance; the motion compensation is used to compensate for displacement errors caused by satellite or platform motion; and the focused imaging is used to generate a high-resolution target image.

3. The system according to claim 1, wherein, The data fusion module achieves accurate registration of multi-view images through a cross-correlation algorithm, and then uses weighted average or principal component analysis methods to fuse the registered images.

4. The system according to claim 1, characterized in that: The error correction module compensates for the reconstructed 3D model. Based on the formula: Δh=k1σ+k2tanθ+k3 / B, the module establishes a functional relationship between the residual error and the noise variance σ, the incident angle θ, and the baseline length B. The compensation coefficients k1, k2, and k3 are fitted by the least squares method, and iterative optimization is performed in combination with actual observation data.

5. The system according to claim 1, characterized in that: The data fusion module achieves accurate registration of multi-view images through a cross-correlation algorithm, ensuring that the error is controlled within 0.1 pixels.

6. A method for generating multi-view images based on SAR satellites, comprising the following steps: Acquire multi-view SAR images of the same target area; The acquired multi-view SAR images are subjected to slant range correction, motion compensation, and focused imaging to generate high-resolution preprocessed images; The preprocessed images are registered and fused to generate a comprehensive dataset containing multi-angle scattering features; Extract interferometric phase information from the comprehensive dataset; Three-dimensional height information of the target region is generated based on the interference phase. Error correction adjustments are made to the three-dimensional height information to ensure the accuracy of the three-dimensional reconstruction; The final target area visualization image is generated based on the corrected and adjusted three-dimensional height information.

7. The method according to claim 6, wherein, The registration and fusion of the preprocessed images includes registration using a cross-correlation algorithm, followed by fusion using a weighted average or principal component analysis method.

8. The method according to claim 6, characterized in that, Residual error compensation for the three-dimensional height model is performed based on the following formula: Δh=k1σ+k2tanθ+k3 / B Where Δh is the residual error, σ is the noise variance, θ is the incident angle, and B is the baseline length; the compensation coefficients k1, k2, and k3 are fitted by the least squares method, and the fitting results are iteratively optimized using actual observation data to obtain the final error-corrected three-dimensional model.

9. The method according to claim 6, characterized in that, The obtained corrected fused data includes the following steps: Cross-correlation algorithm is used to process multi-view images to achieve accurate registration; the registration error is controlled within 0.1 pixels; the registered images are fused using weighted average or principal component analysis to generate a comprehensive dataset.

10. A computer system for implementing the method of any one of claims 6-9, comprising a memory, a processor, and a display device connected thereto, wherein the memory stores computer-executable instructions that, when executed by the processor, cause the system to perform the method of any one of claims 6-9.

Images