摘要:
提出一种基于双目被动视觉的三维人脸识别方法, 该方法采用非接触式的人脸信息采集技术, 利用图像中弱特征检测方法实现双目视觉中的人脸检测与初步视差估计, 运用基于复小波的相位相关技术对人脸表面进行亚像素级小区域匹配, 重建人脸三维点云信息. 通过可调训练次数的神经网络技术实现多层次人脸曲面重建, 并结合人脸2D图像对重构曲面进行仿射归一, 继而迭代地进行特征提取与识别过程. 实验结果表明, 双目视觉方法使人脸信息采集过程友好隐蔽; 在对应点匹配中, 运用复小波的相位相关算法可获得密集的亚像素精度配准点对, 用神经网络方法可正确重建人脸曲面. 识别过程对环境以及人脸位姿表情等鲁棒性强. 该系统成本十分低廉, 适合在许多领域推广应用.
关键词:
-
双目视觉 /
-
三维人脸 /
-
重构 /
-
识别 /
-
神经网络
Abstract:
In this paper, we present a 3D face recognition method based on passive binocular stereo vision. We introduce non-contact face information-collecting technique and use weak feature-detection method on static image to achieve face disparity estimation. Furthermore, we employ phase-correlation in complex wavelet to perform sub-pixel small region matching on the surface of human face, and reconstruct 3D point cloud information of human face. Neural network is applied to obtain multilevel face surface reconstruction. Combining with 2D image of human face, we affine and normalize the reconstructed surfaces, carry out feature extraction and recognition. The experiment shows that this method is robust against surroundings, as well as the position and expression of human face. Moreover, it has high accuracy and quick recognition speed.
下载:
