稻叶瘟染病程度的可见-近红外光谱检测方法  

Determination of rice leaf blast disease level based on visible-near-infrared spectroscopy

作　　者：程术希[1] 邵咏妮[1] 吴迪[1] 何勇[1] 

机构地区：[1]浙江大学生物系统工程与食品科学学院,浙江杭州310029

出　　处：《浙江大学学报（农业与生命科学版）》2011年第3期307-311,共5页Journal of Zhejiang University:Agriculture and Life Sciences

基　　金：浙江省重大科技专项重点农业资助项目(2009C12002);国家农业科技成果转化基金资助项目(2009GB236005);国家自然科学基金资助项目(60802038)

摘　　要：基于可见-近红外光谱技术,并采用偏最小二乘算法对不同水稻稻叶瘟染病程度的叶片进行化学计量学分析,分别建立基于全波段、特征波段和特征波长的稻叶瘟染病程度定量检测模型.结果表明：全波段建模的叶瘟病染病程度检测正确率达到96.7%;通过偏最小二乘算法的回归系数选择5个特征波段,分别为552-558、672-682、719-726、756-768和990-998 nm,基于特征波段的模型正确率也达到了90%,说明该5个特征波段与叶瘟病染病程度有很好的相关性;基于特征波段结果,选择5个特征波长,对叶瘟病染病程度的检测正确率为80%.说明基于可见-近红外光谱技术方法具有较好的预测能力,为稻叶瘟染病程度的快速鉴别提供了一种新方法.A rapid determination of rice leaf blast disease based on visible-near-infrared spectroscopy was proposed.Chemometric analysis was executed on the spectra of the rice leaves with different disease level by using partial least square regression（PLSR）.Three PLSR models were established based on full-range spectra（model 1）,spectra at feature wavebands（model 2） and spectra at feature wavelengths（model 3）.The determination correct rate of the disease detection level was 96.7% for model 1.By using the obtained regression coefficients of PLSR model,five feature wavebands were selected,which were at 552-558,672-682,719-726,756-768 and 990-998 nm.The determination correct rate was 90% for model 2.The result showed that there was a good correlation between the disease detection level and the five selected feature wavebands.Five feature wavelengths were further selected based on the feature wavebands.The determination correct rate was 80% for model 3.It is concluded that the visible-near-infrared spectroscopy gives a good determination result and is a new way to fast determine rice leaf blast disease level.

关 键 词：可见-近红外光谱 偏最小二乘算法 水稻 病变叶片 

分 类 号：S435.111.4[农业科学—农业昆虫与害虫防治]