基于AFSA和PSO融合优化的AdaBoost人脸检测算法.pdfVIP

小型微型计算机系统 2016 年4 月第4 期 Journal of Chinese Computer Systems Vol. 37 No.4 2016 基于 AFSA 和 PSO 融合优化的 AdaBoost 人脸检测算法 任克强，高晓林，谢斌 (江西理工大学信息工程学院，江西赣州 341α)()) E-mail: renkeqiωg@21 摘 要:针对传统 AdaBoost 人脸检测算法存在的不足，提出一种将人工鱼群算法、粒子群优化算法和 AdaBoost 算法相结合的 人脸检测算法.该算法利用人工鱼群算法的最佳寻优特性，弥补粒子群优化算法易陷入局部最优的不足，改善粒子样本的枯竭 和退化;在 AdaBoost 训练框架中扩展了 Haar-like 特征，以排除相关度较低的 Haar-like 人脸样本特征;采用融合优化的 A由­ Boost 算法寻找弱分类器权重系数的最优佳，组合最佳弱分类器，级联形成最终的强分类器.实验结果表明，该算法能够有效提 高检测精确率、降低训练时间，取得了较好的人脸检测效果. 关键词:人脸检削;人工鱼群优化;权重系数;最优弱分类器 中图分类号: TP391 文献标识码:A 文章编号:1仪lO-1220 ( 2016 ) 04-0861 -05 AdaBoost Face Detection Algorithm Based on Fusion Optimization of AFSA and PSO REN Ke-qiang ,GAO Xiao-lin ,XIE Bin ( School of lnfonnation Engineering ,Jiangxi University of Science and Technology ,Ganzhou 341仪泊，China) Abstract :Airning at the shortcornings of 由E 位aditional AdaBoost face detection algorithm ，也is article proposes an algorithm of fa∞ detection based on AFSA ,PSO and AdaBoost. The algorithm uses 由e optimal searching characteristics of AFSA to mak:e up for the shortage of PSO easy to fall into local optirnization ,and improves p缸ticle s缸nple depletion and degradation. In order to elirninate the low correlation Haar-like face s缸nple features ,Haar-like features are extended in the AdaBoost 位aining framework. AdaBoost algo- rithm of fusion optirnization is used to find the optimal value of the weak: classifier weighting coefficient ,and combines the optimal weak: classifier to form the final strong classifier. The experirnental resu1ts show that the algorithm can effectively enhance the detection accuracy rate and reduce training time ,and achieves better result on face det巳ction. Key words: fac巳 detection j article swarm optirnization j weight coefficient j optimal weak: classifier