3D Face Recognition Using Geometric Information

After receiving almost 30 years research, image-based face recognition systems perform well under controlled conditions. However, change of illumination, pose and expression still challenges image-based face recognition. According to the development of 3D acquisition techniques, more and more people turn to utilize 3d facial model to recognize faces.In this paper, there are four contributions which we have been made:1) Provided a survey of 3D face recognitionWe classify 3D face recognition algorithms to three classes according to its representation:by point cloud and mesh structure; by range images; by feature sets. We also give a discussion of face recognition using mulit-model (2D+3D) information and 3D face recognition with the presentation of expression.2) Investigation of 3D face recognition using facial geometric informationIn this work, we utilize distances (including Euclidean distance and geodesic distance) among facial landmarks to represent the 3D face and recognize, we also investigate the performance of using ratios among distances to 3D face recognition.3) Proposed a coarse to fine 3D face recognition using facial geometric informationIn this work, we present a 3D face recognition method which utilizes 3D facial geometric features. Our recognition process takes two steps. In the first step, a number of candidates are selected from the gallery according to the geometric features we extracted. In the second step, ICP algorithm is performed to match the probe with the candidates to make the final decision. We use CASIA 3D face database to evaluate our system.4) Multi-experiments have been carried out;We have designed several experiments using Bosphorus 3D face database and CASIA 3D face database to verify the performance of the algorithms we have proposed. Results show that they work well.