Research On Reconstruction Method Of Panoramic Image Of Building Layout

The reconstruction of panoramic image of building layout provides necessary prior information for enemy reconnaissance,public security law enforcement,disaster relief and other fields.The electromagnetic signal emitted by the Ultra-Wideband Through-Wall Radar detection system is reflected by the wall,and the building layout inversion reconstruction is realized by receiving the wall echo with information and processing the echo signal with relevant imaging algorithm.There will appear refraction,reflection,higher refraction and higher transmission when the electromagnetic wave penetrates through the wall,resulting in displacement,multi-path false image and other problems in the rear wall.Although some existing image processing algorithms for building layout are relatively mature,they are all based on simple buildings.For relatively complex scenes with multiple targets inside a building scene,the recovery effect is relatively poor.This thesis takes the complex scene with multiple targets inside the building as the research starting point,aiming at the influence of indoor targets on wall imaging,the problem of false images in restoring wall imaging and the fusion of multiple perspectives,conducting the following aspects:1?With the change of velocity and propagation direction when electromagnetic waves penetrate the wall,the problem of positional displacement occurs in the rear wall.The approximate compensation method for the wall is given based on the echo model.At the same time,a fast low-rank and sparse decomposition method is proposed due to the fact that the internal targets,corners and clutter have a great influence on the wall echo.The proposed method solves the low-rank and sparse decomposition by fast iterative soft threshold,which can effectively eliminate the internal targets,corners and clutter of the building and extract the wall echo.2?Although the low-rank and sparse decomposition method can effectively separate the internal targets and the corners from the wall,there still exists noise and false image in the reconstructed wall.Thus,a sparse reconstruction method based on total variation is proposed.By introducing auxiliary variables,the proposed method transforms the gradient constraint problem into augmented extremum problems,decomposing the complex large-scale problem into multiple sub-problems by the alternating direction method.By applying non-monotone linear search,the objective function converges quickly with a larger step size,eliminating false images and residual noise,and efficiently reconstructing clear wall images.3?With the problem that the wall image under single-view detection can not reflect the complete building internal information,the two-layer fusion method under multi-aspect detection and the sparse representation fusion based on dictionary training are proposed.The two-layer fusion method firstly performs M-N-K detection fusion on the input single-view image to eliminate part of the independent noise and to smooth the edge of the wall.Then,improved fuzzy logic fusion is carried out,the two-layer fusion is completed by three steps of fuzzification,fuzzy fusion and defuzzification.While in the dictionary-based sparse representation fusion algorithm,the K-SVD algorithm is used to update the redundancy dictionary column by column.Then,the coefficient is restored by the OMP algorithm,and finally the sparse coefficient is fused by the weighted average rule.On the basis of the sparse representation fusion,the image is further smoothed by the morphological opening operation,and the connected domain detection is applied to obtain the wall contour,removing the redundant boundary points and enhancing the image contrast.Simulation and experimental results show that both the two methods simply and effectively reconstruct the panoramic image of the building layout.