通过人工田间诱发不同等级条锈病,在不同生育期测定了36条感染不同严重程度条锈病的冬小麦冠层光谱及相应叶片的生理生化参量。对测定的冬小麦红边一阶微分光谱进行分析,发现随着病情严重度的增加,红边一阶微分的前峰(700nm附近)越来越明显,后峰(约在725—740nm)越来越不突出,以红边一阶微分的双峰特征随病情指数的变化为基础,设计了两个新型的红边参数:DSr和Ar,它们可以分别用来描述红边一阶微分光谱曲线的陡峭度和不对称性,与其他常用的红边参数(红边位置、红边一阶微分最大值,红边一阶微分所包围面积)相比,新参数反演病情严重度的精度更高。

Winterwheat(Triticum aestivum L.) isone of themost mi portantcrops inChina. Severe epidemic ofyellow rust is the main reason ofwinter wheat yield loss and quality degradation. Hyperspectral remote sensing, acquiring mi ages in narrow and continuous spectral bands, has been recognized as a reliablemethod for crop health monitoring. In researches related to hyperspectral data analysis, the region of the red-near infrared(NIR) transition has been shown to have high information content for vegetation status. This region is generally referred to as the“red edge”. It represents the region of abrupt change in leaf reflectance between 680 and 780nm caused by the combined effects ofstrong chlorophyllabsorption in the redwavelengths and high reflectance in theNIRwavelengthsdue to leaf internal scattering. In this article, the firstderivative ofred edge spectra(Dred) was analyzed forwinterwheatwith different stripe rustdisease severity. Results indicated Dred\nhad characteristic ofmulti-peak; fordiseasedwheat, the frontpeak(700nm) wasmore dominantand the back peak(725—735nm) wasmore flat; red edge position(REP) shifted towards shorterwavelength; bothmaxmi um value ofDred(Dr) and sum of Dred(Sr) decreased. Two new red-edge indices(DSr, Ar) were proposed based on the difference in the shape ofDred between healthy and diseasedwheat. DSr, an index describing thinness ofDred, was defined by the ratio ofDrandSr. When winterwheat is healthy(diseased), Dred is very steep (flat) and henceDSris very large(small).Ar, an index describing asymmetry ofDred, was defined as(S2-\nS1) /(S2+S1),whereS1andS2was the sum of Dred in 680—700 nm and 701—775 nm respectively, which corresponded to the spectral region adjacent to front peak and back peak. When winter wheat is healthy (diseased), the difference betweenS2andS1is very significant(non-significant) and henceAris very large (small). Compared to other conventional red edge indices,DSrandArcan predict stripe rustdisease severity more accurately.