Design And Implementation Of Video Capture And Face Detection System Based On Directshow

With the development of human-computer interaction, face detection, an important part of face recognition, has got more and more attention. Face detection is searching faces in several images, and locating and marking the location of the face. For face d etection, getting images of the need to detect firstly. It illustrates the necessity of a vi deo or image acquisition system. With the rapid development of multimedia technology,multimedia with its own huge advantage received attention and application increasingl y, its rapid development has gradually changed the traditional way of working life. Wi th the rapid development of Internet and intelligent application, many security problems has come, and authentication in various occasions has been inevitable. Face recognitio n is a kind of effective method of character authentication, which has made face detec tion is getting more attention from scholars, so as to extend as an independent researc h direction.The main research content of this article is a video capture system based on Dire ctshow and face detection and extraction based on skin-color information and template matching. In Directshow interface, AllocCapFile is used to setting file size. ControlStre am controls the start and end time. SetOutputFileName sets the save path and type of file. Find Interface collects the function of the filter in link. Through these methods, a video capture system is constituted. The video capture system based on Directshow pre view and capture video from many cameras that connected to the PC. The collected vi deo is processing into images frame by frame. And then the images will be put into f ace detection and extraction based skin-color information and template matching. First of all, trough converting images to YCbCr color space, getting skin-color gaussian mo del. Then getting skin likelihood figure by calculating the likelihood skin. Binarizating the skin likelihood figure using the method of dynamic access threshold to achieve the goal of the skin-color area partition. Decreasing the noise and selecting in the candid ate regions using corrosion expansion、euler number and length-width ratio. At last, co mpleting face detection by the method of template matching. The face region detected will be circled and extracted to forming a series of face region images.Video capture system based on Directshow could run steadily, and collected video is clear and smooth. The system has general versatility and expansibility. Experiments of single and multi faces image have finished. The results show that the correct detec tion rate can reach 97%.