基于自适应人脸切割的三维人脸识别算法

3D face recognition algorithm based on adaptive face cutting

为克服表情变化对人脸识别的影响,提出了一种基于自适应人脸切割的三维人脸识别算法.首先,采用一种自动预处理技术来去除离群点、填补孔洞和归一化姿态,以提高三维人脸数据的质量;其次,通过简化meshSIFT特征的规范化方向并加入形状直径函数描述符,讨论了方向分配和特征描述符的设计问题,改进了meshSIFT特征;最后,通过运用字典构造、压缩与自适应区域切割稀疏分类,提出了一种基于多任务稀疏表示分类最小残差和的自适应人脸切割算法.FRGC v2.0人脸数据库上的实验分析结果表明,所提算法对三维人脸识别具有较高的识别率.

In order to overcome the effects of expression variation on face recognition,a three-dimensional（ 3D） face recognition algorithm based on adaptive face cutting is proposed. First,an automatic preprocessing technique is used to remove the outliers,fill the holes and normalize the pose to improve the quality of 3D facial data. Secondly,by simplifying the canonical orientation of the mesh scale invariant feature transform（ mesh SIFT） as well as jointing the shape diameter function（ SDF） descriptor,the direction assignment and the feature descriptor are discussed and the mesh SIFT is improved.Finally,by using the dictionary calculation,compression and adaptive region cutting sparse classification,an adaptive face cutting algorithm based on the minimum residual sum associated to the multitask sparse representation classification is proposed. The experimental results on the FRGC v2. 0 face database indicate that the proposed algorithm has a high recognition rate for 3D face recognition.