中文摘要：

三维人脸因其对光照，装扮变化的不敏感成为人脸分析和识别发展的新趋势。但极易受表情变化的影响成为三维人脸研究发展的瓶颈。针对此关键问题提出一种新的不受表情变化影响的三维人脸分析方法：1）计算三维人脸曲面的内蕴特征测地线距离，提取等距测地线来描述三维人脸曲面特征；2）利用改进的SIFT算法进行特征匹配。在具有不同表情的三维人脸深度图上的实验结果表明，该方法无需复杂的注册，表情的鲁棒性和识别效果优于传统的三维人脸分析方法。

英文摘要：

3 D face analysis and recognition are new trend in the field of face analysis and recognition owning to its insensitivity to variational illumination and dress. However, 3D face is very susceptible to expression change, and it has became a bottleneck of 3D face analysis and recognition. A novel method of 3D face analysis, which is invar- iable to changeable expression, is introduced. Firstly, the geodesic distances on facial surfaces as inherent features are computed, and ISO-geodesic lines to represent 3D facial surfaces are extracted. Then, the modified SIF3＂ algo- rithm is performed to extract the feature key points of the new facial representations for matching. The experiment results on 3D facial depth images show that, compared with any other classical 3D facial analysis and recognition al- gorithms, the proposed method does not need costly registration and is robuster to facial expression change.