Face Detection And Implementation On DSP

With the continuous improvement of people's awareness for security, video surveillance has been widely used in various fields. Either industrial production or daily life, video surveillance is more and more popular with its intuitive, easy, real-time characteristics. However, most of today's video surveillance still needs the operator's monitoring, in order to liberate operator from heavy work of monitoring, a new digital, networked, intelligent video surveillance system come into being. Digital image processing technology makes probability for intelligent video surveillance, and will bring more applications in this field.Human face is the most important part in video surveillance, and a very promising way of authentication. This paper mainly focuses on the implementation of DSP-based face detection. After studying the current mainstream of face detection algorithms and their principles, we find out that the AdaBoost algorithm is more extensive in the field of computer simulation and practical application. So we decide to use AdaBoost face detection algorithm as the basic algorithm, and train samples for classifiers. Then the facial regions are obtained by cascade classifier in the face detection. TMS320DM6437 is one kind of DaVinci digital video image processing DSP produced by TI Incorporated. The DM6437EVM contains DSP subsystem, video processing subsystem and related peripherals. After optimizing AdaBoost-based face detection software program and porting the code to the DSP EVM, the face detect program can be run on the DSP system. It turns out that this real-time detection greatly reduces detection time, and improves detection efficiency. In this way, the system can provide an effective practical application for human face detection and recognition of the relevant video surveillance.