棉花棉叶螨叶片遥感监测技术研究

Remote sensing monitoring technique on cotton leaves infected by spider mites

通过使用便携式光谱仪对新疆石河子地区不同程度棉叶螨为害后的叶片光谱进行测量,同步进行理化参数测定,定性和定量地分析了棉叶螨为害后叶片叶绿素a、叶绿素b、叶绿素（a＋b）、类胡萝卜素、厚度、全氮含量和含水量的变化和光谱反射特性,建立了棉叶螨叶片光谱诊断模型.结果表明：棉叶螨为害后的叶片所有理化参数较正常叶片低,叶绿素a变化幅度最大（88.8%）,叶绿素（a＋b）与棉叶螨的相关性最高（r=-0.924）.棉叶螨叶片光谱反射率大小在近红外（750~860nm除外）和短波红外波段均随严重度的增加而增加,且可见光波段650nm附近有一个独特的反射峰,但在550nm附近低于正常叶片.605~724nm可作为棉花棉叶螨叶片光谱识别的敏感波段,706nm和758nm为最佳波段.4个估测参数建立的诊断模型均达到极显著相关水平,其中R758nm/R706nm建立的诊断模型检验R2最高（0.823）,RE最小（0.351）,可推荐为棉花棉叶螨叶片光谱识别的最佳模型.利用高光谱遥感可实现对棉花棉叶螨叶片的有效识别,可为棉花棉叶螨大面积监测提供借鉴和参考.

Application of hyperspectral remote sensing to monitor cotton leaves infected by spider mites can provide reference for further monitoring at large scale.In this experiment,aportable spectrometer was used to measure spectra,Chl a,Chl b,Chl（a＋b）,carotenoid,leaf thickness,nitrogen content,water content of cotton leaves infected by spider mites in Shihezi area of Xinjiang,the physical and chemical parameters and spectral reflection characteristics were analyzed qualitatively and quantitatively,and spectral diagnosis models were established for cotton leaves infected by spider mites.The results showed that all the physical and chemical parameters of cotton leaves infected by spider mites were lower than those of normal leaves,chlorophyll a had the highest change range（88.8%）,chlorophyll（a＋b）had the highest correlation with severity level（r=0.924）,spectral reflectance of cotton leaves infected by spider mites in near infrared wave band（except 750~860nm）and short wave infrared wave band increased with the increasing of se-verity,and there was a unique reflection peak near 650 nm,but lower than normal leaf.605~724nm wave band could be used as sensitive bands and the 706 nm and 758 nm were the optimal bands for spectrum recognition infected spider mites.Diagnosis models form four spectra parameters all reached the significant correlation level,the diagnosis model of R758nm/R706 nm had the highest R2（0.823）,and the lowest RE（0.351）,and it could be recommended to identify the best model for cotton leaf infected spider mites.In conclusion,it could effectively recognize cotton leaves infected by spider mites using hyperspectral remote sensing.