Face Detection of Video Image Sequence Based on Human Characteristics

Abstract

The study uses DSP as the core and develops agile tracking and surveillance system. It includes overall design of the system, construction of the hardware and compilation of the software. The system hardware consists of DC, color camera, cradle head, SEED-VPM642 DSP development board or pan-tilt-zoom control system. The paper analyzes common algorithms of motion target detection. According to the characteristics of the acquired video images and the real-time requirements of the system, based on adjacent frame difference, the paper proposes continuous three-frame image difference method to detect moving objects and uses centroid algorithm as the core algorithm of human motion tracking. The paper uses the character of human complexion and applies face detection algorithm based on complexion, which can rapidly and accurately get the cleat face information of moving people. And the study selects the sum of the absolute value of gray difference as automatic focusing evaluation function of the design. Optimizing common focusing mountain-climb searching algorithm makes the system focus rapid and reliable, which can achieve the objective of automatic detection and trace on moving body and human head location.