关键词: 人脸检测, AdaBoost算法, 双阈值, 粒子群寻优, 弱分类器

Abstract: For the poor performance and the long training time of the traditional AdaBoost classification algorithms，a new AdaBoost classification algorithm was proposed based on Particle Swarm Optimization（PSO）. Firstly，a new weak classifier structure was established by replacing the original single threshold weak classifier with a double threshold weak classifier. Secondly，searching the optimal feature and two optimal thresholds by PSO were used ro replaece the traditional enumeration search method. Finally，all weak classifiers were combined into a strong classifiers. The theoretical analysis and simulation results show that the average training time of the method based on PSO is reduced to 1/57 of the traditional enumeration method，while the performance of the two methods is equivalent. Experimental results show that the AdaBoost algorithm based on PSO can improve the training efficiency of classifier.

Key words: face detection, AdaBoost algorithm, double threshold, Particle Swarm Optimization (PSO), weak classifier