3d Face Recognition Algorithm Based On Contour Curve Features

Face recognition has received much attention in the area of biometric authentication, because face merges abundant and distinctive nature information of a person, and the information is robust and recognizable.2D face recognition technologies have been studied extensively over the last decades, and applied in many fields. However, they are readily degraded by the environment illumination conditions, and the poses and expressions of a face, making its application more difficult. Instead, 3D face recognition technology is more robust to the illumination perturbation, and the variations of poses and face expressions. Accordingly, it has become a very active research topic in the field of face recognition.In this thesis, a 3D face recognition method is proposed based on facial contour features. It mainly introduces the methods of face feature extraction, and classifier design. The main contribution of this paper is given as follows:First, a method to extract facial contour curve features is proposed based on FCM cluster algorithm. It mainly solves two questions:"What contour curves are extracted?" and "How many contour curves are extracted?". Compared with fuzzy equivalence relation cluster and interval extraction algorithms, the contour curves extracted by FCM cluster algorithm achieve better face recognition performance.Second, based on the Shanon's entropy theory, the left or right side of a face is chosen in the face recognition in terms of their entropy values, as usually the left and right face of a person are symmetric. Experimental results show that the computational complexity is reduced greatly based on our proposed approach, enhancing the speed of face recognition.Third, a new classifier is proposed based on Manhattan distance algorithm, which improves the classification accuracy with the data of face features. Compared with the other classifiers, the modified Manhattan distance algorithm shows better performance.To sum up, this research attempts to develop a more efficient framework for 3D face recognition from the practical point of view. The methods and algorithms proposed outperform the previous approaches in recognition accuracy with less computational complexity, and they are robust to face expressions and illumination conditions.