基于遗传组合网络的玉米品种识别方法  

Composite GA-RBF Networks for Corn Variety Identification

作　　者：孙钟雷[1] 万鹏[2] 龙长江[2] 

机构地区：[1]长江师范学院生命科学与技术学院,重庆408100 [2]华中农业大学工学院,武汉430070

出　　处：《农机化研究》2012年第8期102-106,共5页Journal of Agricultural Mechanization Research

基　　金：重庆市教委科学技术研究项目(KJ101302);华中农业大学引进人才科研启动基金项目(52204-09079);国家自然科学基金项目(61007058);中央高校基本科研业务费专项资金资助项目(2010QC006)

摘　　要：为了提高计算机视觉技术对玉米品种的识别率,研究基于遗传组合神经网络的玉米品种识别方法。利用自制的动态计算机视觉系统,获取了4种具有代表性的西南地区杂交玉米种子图像,提取了8个形态特征参数和6个颜色特征参数。针对玉米品种识别特点,设计了由4个子网络组成的遗传算法优化的径向基函数组合网络,并优选了特征参数组合,优化了网络参数。经验证对比试验表明,本方法识别率可达95%,优于其它识别方法,切实提高了玉米品种识别率。In order to improve corn variety identification detected by computer vision technology,a recognition method was studied,which is based composite GA-RBF networks.Four representational southwest hybrid corn images were obtained,eight shape feature parameters and six color feature parameters were extracted by dynamic home-made computer vision technology.To corn variety character,composite GA-RBF networks were designed,which made up of four branch networks.Composite feature parameter and network parameter were optimized.After validation and contrast,results show that the recognition rate of the method is up to 95%,better than others and the method can improve corn variety identification earnestly.

关 键 词：玉米 品种识别 遗传算法 组合神经网络 

分 类 号：S126[农业科学—农业基础科学] TP391.41[自动化与计算机技术—计算机应用技术]