Font Size: a A A

3D Face Recognition Using Ifnormation Fusion

Posted on:2012-12-28Degree:MasterType:Thesis
Country:ChinaCandidate:C C ZhangFull Text:PDF
GTID:2218330338963710Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
As a booming technology, face recognition has been studied for many years and is expected to be widely used in daily verification and identification systems, communication systems, public security systems, and law enforcement systems and so on. Current two-dimensional face recognition techniques and comprehensive state of art of face recognition algorithms have been presented by Zhao. It still faces some challenges, such as pose and illumination variations, facial expressions and so on. In order to solve these problems,3D face recognition algorithms are being developed. This paper addresses the problem of face 3D face recognition under varying head poses and different lighting conditions.The task of face recognition involves two procedures:feature extraction and feature matching. It is now widely accepted that 3D face recognition outperforms 2D face recognition. However, there are still some challenges to 3D face recognition, such as accuracy improvement which means to enhance the accuracy of 3D face recognition and efficiency assurance which means to decrease memory space requirement and computational costs. This paper aims to solve these two problems.Various face features provide varying discriminative information and contribute to distinct discriminative capabilities. They have different contributions in face recognition. Single features can provide the limited discriminative information and local applicability. Face recognition based on a single face feature cannot achieve satisfactory performance. Thus, it is a reasonable approach to fuse more kinds of face features. Information from multiple sources can be consolidated at distinct levels, including feature fusion, match score fusion, and decision fusion. At feature level, the feature sets extracted from multiple data sources can be combined to create a new feature set to represent the individual. At match score level, different feature-matching outputs are fused to generate a single scalar score for classification. Feature fusion and match score fusion and a proposed two-level are used in this paper to improve the recognition accuracy.This paper proposes a novel 3D face recognition algorithm using multi-level multi-feature fusions. A new face representation method named average edge image is proposed in addition to traditional ones such as maximal principal curvature image and range image. In the matching process stage, a new weight calculation algorithm based on the sum rule is presented for feature fusion and match score fusion. Depending on the complementary characteristic of feature fusion and match score fusion, a combination of them named two-level fusion is proposed. In order to achieve efficiency, Mesh simplification is utilized for data reduction. Experiments are conducted using our own 3D database consisting of nearly 400 samples. Through performance analysis of those three kinds of facial feature images, we can draw that recognition using range image achieves the best performance, and the recognition method PCA+LDA outperforms PCA only. Through performance analysis of the new weight calculation method and those three kinds of fusion algorithms, we can conclude that the new weight calculation method improves the recognition accuracy and the two-level fusion algorithm performs better than feature fusion and match score fusion. Based on the effect analysis of mesh simplification, we can draw that mesh simplification does not only improve recognition accuracy, it also reduces the storage requirement and computational cost.Another contribution of this paper is that an automated face recognition system is developed, which can be used for both 2D face recognition and 3D face recognition. For 2D face recognition, it is a real-time system. Given an ordinary 2D camera, we can obtain facial images immediately and implement recognition. In addition, a scattered data processing system is developed to construct databases of facial feature images from 3D raw data as well as to do some image processing for 3D face recognition.
Keywords/Search Tags:3D face recognition, multi-feature multi-level fusion, feature fusion, match score fusion, two-level fusion
PDF Full Text Request
Related items