3D Face Recognition Based On The Depth Information

Human face is an important biometric to the identity confirmation. As a newly emerging field of the biological recognition technology, face recognition has a theoretical and application significance. It has been a research hot spot in the field of biological recognition in the recent decades. All of the traditional human face recognition algorithms are studied and designed upon the two-dimensional images and their functions are affected by the light, human face gestures and the changes of the expression. As a result, these factors become a huge obstacle to the development of the human face recognition.With the development of the computer vision in the recent years, it has become a possibility to get the real-time three-dimensional information of the target through the camera. So more and more researchers started to focus on the field of the three-dimensional human face recognition expecting that can perfectly solve the problems which can’t be solved upon the two-dimensional images, and intend to optimize the effect of the human recognition and expand the application field of the technology. Under this trend, the paper studied the 3D human face recognition technology in detail. The main work and the innovation points are as follows:(1)After a further study on the methods of obtaining the depth information and comparing the experiment results, the paper determined to obtain the depth information through the laser speckles which is suitable for the application of the human face recognition.(2)An algorithm to optimizes the depth information of human face recognition based on both the fixed stadia and non-fixed stadia was proposed. The algorithm based on the fixed stadia uses the stadia changing method and anther one uses the Media filter to remove the noise、column scanning method to remove holes and histogram equalization to enhance the contrast. The result of experiment shows that the optimized depth information adapts to the face detection.(3)A rapid human face detection algorithm based on the depth information was proposed which can be applied to different scenarios and has a fast detection speed and good robustness. First, the algorithm detects the position of the noise in the depth image. Then, using the skin model to find the area of the face and contour detection to preclude the false target.(4)Studied the face matching method based on the SURF algorithm. The difference between the SURF and SIFT algorithm was analyzed and the possibility of using the SURF algorithm in face recognition area was proved by the experiment.