Research On The Key Technologies For The Projector-Camera Interaction

The projector-camera interaction is a branch of the human-computer interaction research in this paper, This new interactive mode sets of advanced projection display technology, computer vision, pattern recognition and artificial intelligence technologies. At present, there are still some key technical problems have not been solved. For example, the mobility and registration of the system, the description and recognition of hand gestures, precise automatic tracking of hands. In this paper, methods to solve three of the problems are introduced mainly.First of all, analyzed the camera imaging model and corresponding relationship between the projector and camera, reconstructed the projection screen and added basis points to assistant system registration. This method can achieved the purpose of stable output image with less number of points.Used the "user-centric" theory of human-computer interaction, analyzed the cognitive behavior model and its application in gesture interaction, then analyzed the gesture recognition based on the HMM method. We used the shape feature description and matching method to studied the origin and definition of interaction gesture, used the Freeman chain code with suitable correction to described gesture shapes. Considering the gesture interaction in the scene, Interactive gestures including element gesture, and some of the more complex gestures, we disintegrated the complex interaction gesture into element combination. First, described and code the element gesture, then describe the complex continuous gesture from spatial position and sequential logic relation.After the completion of the gesture description and definition, this paper also analyzed the advantage of YCbCr color space in hand tracking recognition, and the improved ellipse skin clustering model in hand segmentation.Because the hand motion is continuous, to study "flexible" gesture recognition method in a certain extent and range has practical significance. Therefore, this paper adopted the continuous deformation theory to tracking hand path.Then, according to the basic characteristics of the physiological characteristics and the camera imaging, we divided the characteristics of hand into U state feature point, V state feature point and CL line. Used the multi-scale and multi structure response model, respectively to extract useful information from these three types of features.During the gesture matching phase, we used the modified Hausdorff distance (MHD), abstracted and matched the element gesture as point, line and curve. Then, matched the continuous gesture from three aspects of the numbers, spatial position and sequential logic relation, taken the highest similarity gesture as the recognition results.Finally, built a projection-camera interactive system and designed the corresponding experiment. The experimental results show that this system has the following characteristics:this system can be moved and automatic registration, the complex gestures are decomposed into the combination element gesture, thus reducing the description difficulty, divided the characteristics of gesture and used the multi-scale and multi-point structure response model increased the recognition accuracy.