Deep Learning Algorithm Based on Adaptive Threshold Segmentation: Innovative Application of BIM Technology in Digital Preservation of Chinese Palatial Architecture

This paper introduces a novel deep learning algorithm for the digital preservation of Chinese palatial architecture. Our approach is based on adaptive threshold segmentation and 3D Convolutional Neural Networks (3D CNN), combined with Physics-based Rendering Techniques. This method effectively handles the complex structures and details of palatial architecture images, achieving high-precision image segmentation and 3D reconstruction. Experimental results demonstrate that our algorithm excels in preserving architectural details and enhancing visual realism, especially in rendering effects. Despite increased resource consumption, the algorithm shows great potential in the application of digital preservation of palatial architecture. Future research will focus on further optimizing the efficiency of the algorithm and enhancing the realism of the model.