基于组合模拟波段的油菜菌核病早期诊断方法研究  

Study on the Early Detection of Sclerotinia of Brassica Napus Based on Combinational-Stimulated Bands

作　　者：刘飞[1] 冯雷[1] 楼兵干[2] 孙光明[1] 王连平[3] 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 [2]浙江大学生物技术研究所,浙江杭州310029 [3]浙江省农业科学院植物保护与微生物研究所,浙江杭州310021

出　　处：《光谱学与光谱分析》2010年第7期1934-1938,共5页Spectroscopy and Spectral Analysis

基　　金：国家高技术研究发展计划(863计划)项目(2007AA10Z210);国家自然科学基金项目(60605011;60802038);浙江省重大科技专项重点农业项目(2009C12002);浙江省研究生创新科研项目(YK2008014);中央高校基本科研业务费专项资金项目资助

摘　　要：应用组合模拟波段建立的线性和非线性判别模型实现了油菜菌核病的早期诊断。采集油菜健康叶片80个,菌核病染病叶片100个,采用预处理算法与连续投影算法(SPA)相结合提取组合模拟波段,分别建立偏最小二乘法(PLS)、多元线性回归(MLR)和最小二乘-支持向量机(LS-SVM)模型。通过比较,最优PLS判别的预处理分别为直接正交信号校正(DOSC)、De-trending和原始反射光谱(Raw),准确率分别为100%,95.7%和95.7%。应用组合模拟波段的最优线性模型为SPA-MLR(DOSC)和SPA-PLS(DOSC)模型,准确率均为100%,基于DOSC、De-trending和Raw组合模拟波段的SPA-LS-SVM模型的判别准确率均为100%。结果表明,基于组合模拟波段进行油菜菌核病早期诊断是可行的,为油菜菌核病的早期诊断及病害监测仪器的开发提供了方法和依据。The combinational-stimulated bands were used to develop linear and nonlinear calibrations for the early detection of sclerotinia of oilseed rape （Brassica napus L. ）. Eighty healthy and 100 Sclerotinia leaf samples were scanned, and different preprocessing methods combined with successive projections algorithm （SPA） were applied to develop partial least squares （PLS） diseriminant models, multiple linear regression （MLR） and least squares-support vector machine （LS-SVM） models. The results indicated that the optimal full-spectrum PLS model was achieved by direct orthogonal signal correction （DOSC）, then De-trending and Raw spectra with correct recognition ratio of 100%, 95.7% and 95. 7%, respectively. When using combinational-stimulated bands, the optimal linear models were SPA-MLR （I）OSC） and SPA-PLS （DOSC） with correct recognition ratio of 100%. All SPA-LS-SVM models using DOSC, De-trending and Raw spectra achieved perfect results with recognition of 100%. The overall results demonstrated that it was feasible to use combinational-stimulated bands for the early detection of Sclerotinia of oilseed rape, and DOSC-SPA was a powerful way for informative wavelength selection. This method supplied a new approach to the early detection and portable monitoring instrument of sclerotinia.

关 键 词：可见/近红外光谱 油菜菌核病 直接正交信号校正 连续投影算法 最小二乘-支持向量机 

分 类 号：O657.3[理学—分析化学] S435.6[理学—化学]