差分进化在图像特征选择中的应用研究  

Research on the Application of Differential Evolution for Image Feature Selection

作　　者：苗玉杰[1] 

机构地区：[1]中国环境管理干部学院信息工程系,河北秦皇岛066004

出　　处：《科技通报》2013年第8期115-117,共3页Bulletin of Science and Technology

摘　　要：图像特征选择是图像检索至关重要的一个环节,本文将差分进化计算应用于图像特征选择中。针对差分进化计算存在过早收敛的问题对其进行改进,给出了改进算法的交叉、变异、选择操作过程,同时将两种算法应用于图像检索中。利用Corel标准图片库测试算法的性能,结果表明改进算法的运行时间与原算法相当,但其查全率与查准率明显优于原算法。Image feature selection is a crucial link for image retrieval,and the differential evolution is applied to image feature selectiono An improved algorithm is gived because differential evolution has the problem of premature conver- gence.Also crossover, mutation,selection process is gived detaily.Two kinds of algorithms has been used in image retrieval. The per/bnnance of both algorithm is tested using he Corel standard picture library. The results show that the improved ＇algorithm almost has the same running time compared to the original algorithm,but it has better retrieval accuracy.

关 键 词：差分进化 特征选择 交叉变异 查准率 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]