基于扩展对称局部三值模式的三维人脸识别

3D face recognition based on extended symmetric local ternary pattern

本文提出了一种有效的三维人脸识别方法,局部描述子是处理识别问题的最有效的方法之一,使用一种全新的人脸分割方案在三维面部中提取表情不变特征区域,在三维人脸多边形网格上融合扩展中心对称局部二值模式(XCS-LBP)和中心对称尺度不变三值模式(CS-SILTP)提出新的三维局部描述子XCS-SILTP对三维人脸表面法向量差值编码进行纹理描述。采用本文提出的方法在FRGC数据库上进行试验,结果表明该方法对人脸表情变化,局部遮挡和适度的姿态变化都具有较强的鲁棒性,采用XCS-SILTP描述子的算法进行人脸识别是一种切实有效的方法。

In this paper,an effective 3D face recognition method is proposed.Local descriptor is one of the most effective methods to deal with the recognition problem.In this paper,a new face segmentation scheme is used to extract the expression invariant feature in 3D face.A new 3D local descriptor,XCS-SILTP,is proposed to describe the texture on the surface of 3D face by blending the extended centrosymmetric local binary pattern(XC-LBP)and centrosymmetric scale-invariant ternary pattern(CS-SILTP)on the 3D face polygon grid.The method proposed in this paper is tested on FRGC database.The results show that the method has strong robustness to facial expression changes,local occlusion and moderate posture changes,and proves that the XCSSILTP descriptor algorithm is a practical and effective method for face recognition.