Research On Face Detection And Tracking Algorithm

Face detection and tracking is one of the most famous subjects in computer vision and patter recognition, a lot of new methods have been introduced within many years of research on it, and with the help of these methods, the precision and rate of face detection and tracking are greatly developed. This thesis introduces methods both on face detection and tracking based on former research,the results prove that the methods successfully improve the precision of them.This thesis introduces the correlative research of face detection, it sets up a face detection system using the Adaboost algorithm, and introduces a new optimized method. The optimized method which uses interior struct detection and interior struct or face nesting elimination to reduce the rate of mistaken detection effectively. This thesis sets up a face detection system using Visual C++ programming in Windows and it sets up a self-created library with it. This thesis trains both MIT library and self- created library to set up the Adaboost cascade classifier, and it uses lots of computer experiments to prove that the optimized method reduces the rate of mistaken detection effectively both with the MIT library and self-created library. So the optimized method improves the precision of face detection.This thesis introduces the correlative research of face tracking, it sets up a face tracking system using the Camshift algorithm, and introduces a new optimized method. This method uses Adaboost face detection to initialize the templet instead of the manual templet initializing. This thesis uses Visual C++ programming in Windows to set up a face tracking system without people. It introduces a new optimized method to solve the instability of templet initializing using face detection. The optimized method reduces the instability of templet initializing with the help of detecting the changes of templet size and position. So the optimized method improves the precision of face tracking.This thesis which combines the characteristics of Adaboost face detection and Camshift face tracking and the optimized methods of both of them realizes a fast face detecting and tracking system with the help of programming in Windows. It uses lots of computer experiments to prove that the system which combines these new optimized methods improves the precision of both face detection and tracking.