Deformable Model Based Three-Dimensional Face Reconstruction And Recognition

Three-dimensional face reconstruction and three-dimensional face recognition are hot topics in computer vision and pattern recognition.Three-dimensional face reconstruction of a face image has a wide range of applications in face recognition,computer animation.Compared to traditional two-dimensional face recognition,three-dimensional face recognition has advantages in posture and lighting issues.This paper studies a three-dimensional face reconstruction technology of a single photo and three-dimensional face recognition technology,the main work is as follows:Firstly,a 3D face reconstruction algorithm based on redundant sparse deformable model is proposed.1)The scale of face database is extended and the extended database is used as training samples to construct a redundant dictionary under K-SVD algorithm;2)AAM algorithm is used to extract the feature points of the face image 3)Sparse deformable model and redundant dictionary are combined to reconstruct the face shape.4)Triangular mapping is used to add texture to the 3D face.Experimental results show that the proposed algorithm has high reconstruction accuracy and realism,and the textured 3D face is visually close to the original image.Secondly,a 3D face recognition algorithm based on face range image is proposed.1)3D face is aligned to the reference model.Three-dimensional face is transformed into the face range image by using the translation,the scale and the projection;2)the depth features,the LBP features and the simplified Haar features of the face range image are extracted;3)The principal component analysis to is used to map the three features to the same Space,which are fused to obtain enhanced features;4)Support vector machine is used to train and classify features.Experimental results show that the algorithm has a good recognition effect and robustness to facial expression changes.Finally,a three-dimensional face recognition algorithm based on geodesic stripes is proposed.1)the rigid part of the facial geodesic stripe is extracted as a feature;2)3D Weighted Walkthrough is used as a measurement,and angle weight vector is introduced to measure the spatial relationship between geodesic loops;3)The graphic model is used to measure the difference between three-dimensional faces and "dissimilarity degree" is constructed as the measure of face difference;4)A coarse to fine matching strategy is used,Experimental results show that the proposed algorithm has high recognition accuracy and robustness under the changes of facial expressions.