基于自适应粒子群优化算法的限速标志识别  

The speed limit traffic sign recognition based on improved APSO

作　　者：成健[1] 张重阳[1] 

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《信息技术》2016年第9期5-9,共5页Information Technology

基　　金：国家自然科学资金项目(90820306)

摘　　要：针对经典的粒子群优化算法收敛性能不足的问题,提出了一种自适应粒子群优化算法。该算法对进化过程中的收敛方向进行动态调整,通过学习因子对参数选取过程进行调节,使收敛的初期趋向全局最优,后期趋向局部最优。采用自适应粒子群优化算法对支持向量机模型参数进行优化,并应用于限速标志的识别,实验结果表明,文中基于自适应粒子群优化算法的方法拥有较高的识别率,同时算法收敛性显著高于经典的粒子群优化算法。According to the disadvantage ol the classic P S O algorithm, the A P S O algorithm wasproposed. The algorithm could adjust the direction ol tuning parameters during the evolution, it couldoptimize the parameters ol the support vector machine model accurately and fast by learning factors, itcould approach the overall best result at the beginning while the local best result at the latter part. Thealgorithm could be applied to the experiment of speed limit traffic signs recognition based on H O G -S V M .The result turned out that, the method based on A P S O could achieve high accuracy, meanwhile, theconvergence was significantly higher than the traditional P S O algorithms.

关 键 词：交通标志识别 粒子群优化算法 支持向量机 HOG特征 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]