Research on Denture Feature Matching Point Cloud Data Based on Deep Learning

With the vigorous development of modern science and technology, to effectively solve the shortcomings of traditional oral diagnosis and treatment, digital oral diagnosis and treatment models are based on three-dimensional reconstruction technology to restore the surface information of denture models, making it easier for doctors to accurately diagnose and treat patients' conditions. Deep learning provides new ideas for denture point cloud feature matching. In this paper, the EdgeConv structure is introduced to improve the network for denture point cloud matching. The experimental outcomes demonstrate that the improved network for matching features of denture point clouds performs well in comparison to the traditional ICP algorithm. Compared with the PCRNet network, the MSE_R is reduced by 0.716, and the MSE_t is reduced by 0.064. The improved network has stronger feature extraction capabilities for local features, resulting in point cloud registration outcomes that are closer to the actual point cloud registration. Additionally, the network's performance is more consistent and stable.