Feature Extraction and Matching of 3D Face Model Based on Facial Landmark Detection

Building a high-quality 3D face model is one of the important issues in computer graphics. High-fidelity 3D face model can be obtained through 3D reconstruction methods from measured data, but the reconstructed models are often flawed with missing measurement data. In order to improve the usability and robustness of the 3D face models, repair and optimization of the face model are critically important. The traditional solutions rely on geometric heuristics to complete repairing work. Considering the characteristics of the face, it is possible to utilize the semantic information to improve the quality of repairing process for the three-dimensional face models, thereby improving the accuracy of the repaired models. In this paper, we combine artificial intelligence and computer graphics methods, and proposes a feature extraction and matching of 3D face model method based on facial landmark detection. Our method combines the facial feature recognition algorithm of 2D facial image and the mapping relationship of 3D face mesh model, and realizes the function of semanticsegmentation and extraction of the 3D face model features. It can be used for the next step of 3D face mesh repair research.