A Genetic Algorithm Based Feature Selection Approach for 3D Face Recognition

Research on 3D face recognition [1,9] has been intensified recently due to the significant advances of the 3D imaging technology. Most of the research focuses on the investigation of 3D range data obtained by a 3D scanner. Although 3D capture systems provide highly accurate 3D face information, it is not trivial to process the large amount of facial surface data. For example, it is hard to keep the correspondences among different subjects because their models have different vertices, and a post-processing procedure needs to follow because the range data may contain too much redundant information, which may make the facial shape comparison noise sensitive. In this paper, we propose to use a generic model to construct the 3D facial feature space. Figure 1 shows the scheme of our proposed method. This approach relies on our realistic face modeling technique, by which the individual face model is created using a generic model and two views of a face. The feature space is composed of geometrical structures, the labeled curvature types of each vertex in the individualized model. Some research used statistics approaches, such as PCA, ICA and LDA [11], to form a feature space. Since they mainly rely on feature transformation procedures, which project the original feature set into a more compact set while retain the necessary information, some features that confuse the recognition task may still be chosen. In our approach, the optimized features are selected by using a Genetic Algorithm based approach. This feature selection approach maintains the " good " features that minimizes the inner-class distance and maximizes the intra-class distance. We tested on two sets of databases. One set consists of 105 3D facial models, about 92% rank-four correct recognition rate is achieved. The other set has 387 models, the correct recognition rate is 87.6%. The experimental results show that the features obtained from the 3D individualized model is feasible to classify and identify individual faces. (a) (b) (c) Figure 2: (a) Labeled feature space. (b) Sub-regions. (c) Optimal feature space. Based on our existing work [2], we created a 3D facial model database by modifying a generic facial model to customize each individual face, given a front view and a side view of one face. This approach is based on recovering the structure of selected feature points in the face and then adjusting a generic model using these control points to obtain the individualized 3D facial …