基于ANN和PCA的玉米品种特征分析与识别研究  

Research on the analysis & classification of maize cultivars characteristics based on ANN & PCA

作　　者：熊凯[1] 李向红[2] 李言照[1] 韩仲志[1] 

机构地区：[1]青岛农业大学理学与信息科学学院,山东青岛266109 [2]烟台南山学院计算机系,山东龙口265700

出　　处：《粮油食品科技》2010年第4期1-5,共5页Science and Technology of Cereals,Oils and Foods

基　　金：山东省自然科学基金(2009ZRA02123);山东省科技攻关计划项目(2009GG10009057;2009GG10009005);青岛科技发展计划基金(07-2-1-37-nsh;08-2-1-15-nsh)

摘　　要：研究了一种基于玉米外观形态和颜色特征进行的玉米品种的特征主分量分析及BP神经网络识别方法。采用数码相机获得了11个品种每个品种50粒共550幅图像,然后对各品种对应的籽粒群体图像提取每个籽粒的形态特征8个,颜色特征12个、纹理特征13个,共33个特征参数,采用主分量分析PCA的方法提取其主分量,将这些主分量作为BP神经网络的输入,构建4层神经网络,并分别定义11个玉米品种的二进制编码作为网络的输出,建立特征参数与玉米品种之间的神经网络识别模型。试验结果表明,方法对11个品种550个籽粒的品种检出率为92%以上,得到了较好的识别效果。A kind of maize cultivar classification method was researched which was based on appearance characters by its principal components（PCs） and BP artificial neuron networks（ANN）.550 digital images of 11 maize cultivars were scanned by digital camera.33 characters,i.e.8 morphologic characters,12color characters and 13 texture characters,were extracted.The principal components were extracted by principal component analysis（PCA） which were used as the inputs of BP-ANN,the ANN modal of 4 layers and the recognition modal of the relationship of characteristic parameters and maize cultivars were established.The result indicated that the detection rate is above 92% for 550 digital images of 11 varieties.

关 键 词：玉米品种识别 计算机视觉 主分量分析(PCA) 人工神经网络(ANN) 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论] S126[自动化与计算机技术—计算机科学与技术]