基于区域分割的表情鲁棒三维人脸识别方法

Region-based expression invariant 3D face recognition

为了克服表情变化致使三维人脸识别性能不佳的问题,提出基于鼻尖点区域分割的表情鲁棒三维人脸识别方法。首先,根据表情对人脸影响具有区域性的特点,提出仅依赖鼻尖点的表情不变区域(刚性区域)和表情易变(非刚性区域)划分方法;然后针对表情不变区域和表情易变区域使用不同的特征描述方式并计算匹配相似度;最后将表情不变区域和表情易变的相似度进行加权融合实现最终身份识别。提出方法分别在FRGC v2.0和自建WiseFace表情人脸数据库上达到98.52%和99.01%的rank 1识别率,证明该方法对表情变化具有较强的鲁棒性。

In order to overcome the problems caused by expressions in 3D face recognition,this paper proposed a new method based on nose tip segmentation.First,in order to weaken the effects of facial expression,it proposed a new method that only needed nose tip of dividing faces into rigid and non-rigid regions.Then,considering the discrimination ability of rigid and non-rigid regions,this paper employed different 3D face features of the rigid and non-rigid 3D face regions.At last,the method fused the rigid and non-rigid face features to address the recognition task.Experimental results on the FRGC v2.0 and self-built WiseFace databases show that the proposed method is robust to expression variations.The rank-one face recognitions rates on the two databases are 98.52%and 99.01%separately.