3D Face Recognition Using Corresponding Point Direction Measure and Depth Local Features

2017.11.13

3D Face Recognition Using Corresponding Point Direction Measure and Depth Local Features (IS2010)

对应点方向测量+深度局部特征

步骤：

1. ICP 对齐：Iterative Closet Point (ICP) to align all 3D faces with the first 3D face
2. 去噪：reduce noise,especially the noise which in front of the face, and remove the spikes
3. 鼻尖检测->人脸检测：detect the nose tip point. Once the nose tip is found, crop a region, which is defined by a sphere radius of 100 mm centered at the nose tip
4. CPDM 人脸匹配：Corresponding Point Direction Measure (CPDM) to matching the 3D face with the gallery 3D faces and get the score
5. 构建深度图，获取三种特征：use the region to construct depth image, and get the Gabor feature, LBP feature, principle component of the depth image
6. 融合特征，实现人脸识别：fuse the CPDM result, Gabor feature,LBP feature, and principle component of depth image to finish the recognition

流程图：

3D人脸匹配

• CPDM（ Corresponding Point Direction Measure）

目标：计算具有同样方向的两个对应人脸的点率，computing the rate of points of the two corresponding faces that have the same direction

算法流程：

C is a set while α is a constraint of the angle between two normal vector, use α = 5 in the following experiments

• Gabor feature：保护人脸的特征不变invariant feature of face

在特征提取中对rotation，scale，translation保持不变性

• LBP：纹理描述子

• PCA：保护主要信息

• 匹配结果：融合CPDM和三个特征，取CPDM匹配最大值+主成分匹配的最小差距

ps：

ICP对应点匹配： http://note.youdao.com/noteshare?id=eec9791a3c9817091f042b90799f96e5

数据库：CASIA

结果：ROC、FRR、FAR