Research And Implementation Of The Head Three-Dimensional Motion Simulation Algorithm Based On Monocular Video

The Fields involved of head three-dimensional motion simulation include computer vision, computer graphics, image processing, pattern recognition and artificial intelligence. The research topic is challenging and cross-subject. It has a wide application prospect in the fields including animation, game, intelligent human-computer interaction, intelligent monitoring, video conference, virtual reality and augmented reality.Three dimensions movement analysis is an important part of image processing. It develops on the basis of image processing; mainly studies object motion in space at extracting information of the moving object shape, location and motion parameters. It is an ill-conditioned problem recovering three-dimensional information from monocular video. It is a challenging task how resumes the head three-dimensional motion accurately from monocular video.The head three-dimensional motion simulation algorithm based on feature points'alignment is discussed in this paper. The step function of algorithm is improved, and two methods to obtain the depth information of head are put forward. First of all, extract three face feature point from the video images, constitute a 2D triangle, and then obtain the corresponding three-point spatial location using the three points for affine transformation between a known model and video images, constitute a 3D triangular body parallel to the plane of face, suppose the head movement as a rigid body movement, so the triangular body can simulate the actual head three-dimensional movement. The depth information is difficult in three dimensions movement analysis based on monocular video. Two methods to obtain the depth information of head are putted forward: for obtaining the depth information of head approximately, the first uses the distance between two points, the other obtains the camera parameters with camera calibration algorithm. By simulative testing, the algorithm is simple and effective, the step function improved produce a good effect. The two methods to obtain the depth information have their own strengths, weaknesses and suitable scope.