Face Detection And Tracking

Human face can provide much valuable information in the study of perceptual human-computer interface and compute vision. Although many outstanding algorithms have been created in human face detection and tracking, some serious problems are still exist such as real-time detection, image segmentation, occlusion handling and multiple camera cooperation. In this paper we make some researches on those problems above and several new approaches and algorithms are put forward. Additionally, in order to test the performance of schemes worked out here, a human face detection and tracking system has been built based on multiple camera environment. The main contributions of the paper are summarized as follows:A skin color and gradient based real-time face detection algorithm is presented. After analysis the characters of the human face, we draw a conclusion that the performance both in fast and accuracy can be obtained through suitable character selection and image down sampling. Experiments results show that this algorithm is not sensitive to the head rotation and background changing.A particle filter based face tracking algorithm under complex environment is provide. Particle filter can deal those nonlinear and nongaussian estimation problems, and what's more, it doesn't need the measurement equation which is hard to build in face tracking. Experiments results demonstrate the algorithm is effective under complex situation such as changing background and partial occlusion.In order to solve the multiple camera best view selection problems, not only we present two selection algorithms based on skin ratio and color hist matching, but an improved face detection algorithm based on motion segmentation is presented for static camera.We have developed an auto human face detection and tracking system with VC 6.0 , it's function include face detection, active camera control, computer communication, best view selection, in addition ,it support different video source such as CCD and USB based camera.