Structure-Preserving 3D Reconstruction Algorithm for Traditional Buildings Based on Hybrid Imaging Techniques

This study proposes a novel traditional building structure-preserving 3D reconstruction algorithm based on hybrid imaging technology. The main application scenario of this algorithm is to provide a digital solution for building protection for the deformation and degradation of the traditional buildings caused by natural or human factors. This study first analyzes RGB imaging, thermal imaging, and 3D point cloud as three mainstream imaging technologies, and conducts a preliminary analysis of different modes of image analysis. At the algorithm design level, this study first uses the cross-modal registration algorithm, SIFT feature extraction and neural network density optimization method for comprehensive modeling. This operation achieves accurate alignment of RGB images, thermal imaging images and 3D point cloud data. Then, in the image detail optimization stage. This study uses high-frequency detail enhancement technology based on wavelet transform to enhance details. This operation enhances the texture and edge information of the image. Finally, combined with the partial completion algorithm of the image style transfer. Local style transfer is performed on the target area under the style prior background, thereby completing the restoration and reconstruction of the architectural image. Experimental results show that the proposed algorithm is superior to the traditional algorithm in image restoration quality, and obtains good performance in specific indicators, such as PSNR value of 32.0 dB, SSIM value of 0.85, and LPIPS value of 0.08.