基于高光谱图像技术的番茄叶片氮素营养诊断  

Hyperspectral imaging technology of nitrogen status diagnose for tomato leaves

作　　者：朱文静[1] 毛罕平[1] 周莹 张晓东[1] 

机构地区：[1]江苏大学现代农业装备与技术省部共建教育部/江苏省重点实验室,江苏镇江212013 [2]泰州机电高等职业技术学院,江苏泰州225300

出　　处：《江苏大学学报（自然科学版）》2014年第3期290-294,共5页Journal of Jiangsu University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61075036);江苏高校优势学科建设工程项目(苏政办发〔2011〕6号)

摘　　要：通过提取番茄叶片高光谱图像的灰度、纹理特征将高光谱数据立方体转化成二维特征曲线,再利用特征选择方法 CFS对所有波长进行筛选,确定灰度特征的特征波长:549,669,742,830 nm和纹理特征的特征波长:482,684 nm.每个样本共有12个特征变量,对这12个变量进行主成分分析,提取9个主成分因子作为模型的输入向量,采用支持向量机建立番茄氮素营养水平诊断模型,得到4个梯度氮素水平(N1,N2,N3,N4)番茄叶片的正确识别率为96%,88%,92%,92%.研究结果表明高光谱图像技术对番茄氮素营养水平具有较好的诊断作用.The data cube of hyperspectral image was translated into two-dimension characteristic curve by extracting color and texture features, and all wavelengths were screened by the feature selection method of CFS. The characteristic wavelengths of color feature were determined as 549, 669, 742 and 830 ram,and those of texture feature were 482 and 684 nm. The 12 feature variables were extracted for each tomato leave sample to perform principal component analysis （PCA）. The 9 principal components （PCs） were extracted as input vectors to establish SVC model for the identification of nitrogen status. The experimen- tal results show that the distinct nitrogen status accuracies of N1 ,N2 , N3 and N4 are 96% ,88% ,92% and 92% , respectively. The hyperspectral imaging technology is suitable for nitrogen status diagnose of tomato leaves.

关 键 词：番茄叶片 高光谱图像 灰度特征 纹理特征 特征选择 支持向量机 

分 类 号：S123[农业科学—农业基础科学] TP391.4[自动化与计算机技术—计算机应用技术]