江苏农业学报
江囌農業學報
강소농업학보
JIANGSU JOURNAL OF AGRICULTURAL SCIENCES
2015年
3期
531-537
,共7页
傅里叶变换红外光谱%判别分析%蚕豆%病虫害%诊断
傅裏葉變換紅外光譜%判彆分析%蠶豆%病蟲害%診斷
부리협변환홍외광보%판별분석%잠두%병충해%진단
fourier transform infrared spectroscopy%discriminant analysis%broad bean%disease and pest%di-agnosis
为建立一种基于傅里叶变换红外光谱( FTIR)结合判别分析的蚕豆病虫害诊断方法,以病虫害危害的蚕豆叶片样品FTIR数据为指标,采用逐步判别法,依据Fisher线性判别准则建立判别模型,对样品的病虫害种类和病原物类别进行诊断,比较了不同光谱范围和不同级别光谱数据以及挑选判别指标建立判别函数时5种方法的判别效果。结果表明,基于FTIR数据的判别分析能较好地诊断蚕豆病虫害种类和病原物类别,以波数1800~1200 cm-1的一阶导数光谱数据为判别指标进行诊断时效果较好;采用Unexplained variance逐步判别法对病虫害种类诊断时,正确率相对最高,为93.1%;采用Wilks’ lambda逐步判别法对病原物类别诊断时,正确率为91.8%。FTIR光谱技术与判别分析方法相结合,可为蚕豆病虫害诊断提供一种简便易行的方法。
為建立一種基于傅裏葉變換紅外光譜( FTIR)結閤判彆分析的蠶豆病蟲害診斷方法,以病蟲害危害的蠶豆葉片樣品FTIR數據為指標,採用逐步判彆法,依據Fisher線性判彆準則建立判彆模型,對樣品的病蟲害種類和病原物類彆進行診斷,比較瞭不同光譜範圍和不同級彆光譜數據以及挑選判彆指標建立判彆函數時5種方法的判彆效果。結果錶明,基于FTIR數據的判彆分析能較好地診斷蠶豆病蟲害種類和病原物類彆,以波數1800~1200 cm-1的一階導數光譜數據為判彆指標進行診斷時效果較好;採用Unexplained variance逐步判彆法對病蟲害種類診斷時,正確率相對最高,為93.1%;採用Wilks’ lambda逐步判彆法對病原物類彆診斷時,正確率為91.8%。FTIR光譜技術與判彆分析方法相結閤,可為蠶豆病蟲害診斷提供一種簡便易行的方法。
위건립일충기우부리협변환홍외광보( FTIR)결합판별분석적잠두병충해진단방법,이병충해위해적잠두협편양품FTIR수거위지표,채용축보판별법,의거Fisher선성판별준칙건립판별모형,대양품적병충해충류화병원물유별진행진단,비교료불동광보범위화불동급별광보수거이급도선판별지표건립판별함수시5충방법적판별효과。결과표명,기우FTIR수거적판별분석능교호지진단잠두병충해충류화병원물유별,이파수1800~1200 cm-1적일계도수광보수거위판별지표진행진단시효과교호;채용Unexplained variance축보판별법대병충해충류진단시,정학솔상대최고,위93.1%;채용Wilks’ lambda축보판별법대병원물유별진단시,정학솔위91.8%。FTIR광보기술여판별분석방법상결합,가위잠두병충해진단제공일충간편역행적방법。
To establish a method based on the fourier transform infrared spectroscopy ( FTIR) combined with step-wise discriminant analysis to diagnose the types of diseases and pests of broad bean, the spectral characteristics of the leaf samples attacked by diseases or pests were analyzed using the discriminant model based on the Fisher linear discriminant criterion. The discrimination effectiveness was compared for the range and level of spectral data, as well as the 5 discrimi-nant indexes used for developing discriminant function. The results indicate that that the discriminant analysis based on FT-IR could diagnose the type of diseases and pests and the category of pathogens of broad bean, and the first derivative spectra data in the range of 1 800-1 200 cm-1 should be selected as the discriminant index for best discrimination effectiveness. When dealing with the type identifications of diseases and pests of broad bean, the Unexplained variance method of stepwise discriminant analysis should be used, yielding a 93. 1% accuracy. The Wilks’ lambda method was better for the categori-cal diagnosis of pathogens, yielding a 91. 8% accuracy. As a simple and convenient method, the FTIR combined with discriminant analysis is capable of detecting the disea-ses and pests of broad bean.