Analysis And Design Of Head Motion Posture Sensing

Head pose detection is a research topic in the field of a computer vision, human-computer interaction, intelligent monitoring, fatigue testing. Head pose detection process is to obtain the head position and attitude through the camera and a series of data processing and analysis. The key technology of this paper is around the head motion detection, introduced the research significance of head pose estimation, domestic and foreign research of present situation and the research direction and method of head pose estimation, And simple introduced the structure, advantages and function of OpenCV. This paper mainly discussed the face detection, head posture detection of facial feature points in the system model and simulation and head pose estimatioa Specific work is as follows.First, the camera calibration. Camera calibration is the camera and the real three-dimensional world measurement of contacted bridge, its purpose is to build a 3 d world coordinates and a corresponding relation between the 2 d image coordinates. Camera calibration is a process for the camera parameters, namely make the camera image coordinate system and the object in the space of the process of mapping between the 3 d coordinate system. Based on the study of the basic principles and methods of the previous calibration of the camera calibration, This paper realized the camera calibration based on OpenCV, and provided data for the after 3D pose detection.Second, the face detection. Face detection as the name implies is to detect whether the image contains a human face. Face detection methods used in this paper is based on the improve of the AdaBoost algorithm. The method firstly training a classifier, and then use the trained classifier to detect whether the image contains faces. Face detection is the first step of the human face information processing, is also an indispensable step, it lays the foundation for the later AAM features matching.Again, the face feature extraction. Facial feature extraction is a process of use a The built model to match the face with similar features in the image, and returns the corresponding feature point information. Method for face feature extraction method used in this paper is based on the active appearance model. The method is first to establish the model, get the AAM model can be deformed with parameters. And then, by adjusting the parameters of model, changing model make it to approach the target image, continuously make adjustments in the process, and eventually realize the input location and extract the key points in the image.Finally, use POSIT algorithm to estimate the head posture parameters. The camera internal parameters previously obtained and extracted facial feature input POSIT algorithm, we can estimate the human head posture. Based on the successful detection of the head position and orientation parameters, but also achieve a real-time detection of head posture.Based on the algorithm design, using C and C++ language, Finally realization of a simple real-time head pose detection system. And using the test persons head accuracy and robustness of the system were tested. After analysis, the algorithm has reached the demand requirements, and has a certain reference for the direction of the related research.