The Study Of Hyperspectral Remote Sensing Models Of Fir Major Biochemical Parameters

Hyperspectral Remote Sensing with multi-band and high-resolution characteristics is divided into imaging and non imaging hyperspectral remote sensing.Non imaging hyperspectral remote sensing can record the spectral features of ground feature in detail,not only help understand the properties of aviation or aerospace hyperspectral remote sensing data,but also deeply understand the inherent mechanism of remote sensing imaging.The analysis of spectral features can improve application accuracy of different types of remote sensing data. Identification and classification of vegetation types, estimation of plant chemical composition are the important issue in hyperspectral remote sensing research. At present, the study on spectral characteristics and biochemical component of the crop plant is mature, but less on forest species, and the results are more scattered.In this study, fir which in Huang Fengqiao state-owned forest farm of Hunan Province was selected as interesting object,in the event that the external environmental conditions are basically same,the ground or the air spectral reflectance of canopy was measured by ASD HandHeld Spectrometer in field.Some biochemical parameters including chlorophyll-a content,chlorophyll-b content,total chlorophyll content,carrot content were measured at the same time.ViewSpec Proversion 4.02,SPSS 17.0 (Statistical Package for Social),MATLAB software and Excel software were used to process these datum. Multifactor linear stepwise regression technique and curve fitness analysis,Neural Network Technology were adopted to discuss the correlation between camphor main biochemical parameters with raw hyperspectral reflectance as well as its different transformations,which include its first differential spectral reflectance,hyperspectral characteristic variables,vegetation indices.Then,some estimation models of different biochemical parameters were built based on the results of correlation analysis,finally,the predictive precisions were checked for those chosen models in order to determine the best estimation models for biochemical parameters. The suitable models are following:Chlorophyll-a content:①y=312.033×x508-760.74×x557+0.684 (Note:x508 and x557—the first differential spectral reflectance at 508nm and 557nm)②y=exp (-0.443+8.423×Rg)②y= 0.642exp (8.423×Rg)④neural network models of chlorophyll-a contentChlorophyll-b content:①y= 385.103×x501+212.857×x456+0.421(Note:x50] and x456—the first differential spectral reflectance at501nm and 456nm)②neural network models of chlorophyll-b contentChlorophyll-ab content:①y= 0.291×x395+20.172×x521+0.758 (Note:x395 and x521—the first differential spectral reflectance at 395nm and521nm)②y=exp(-0.085+7.962×Rg)②y=0.919exp(7.962×Rg)④neural network models of chlorophyll-ab content Carrot content:neural network models of carrot contentIn a word, by observating fir spectral and processing data, find out biochemical parameters with a significant effect for the spectral characteristics of fir and hyperspectral remote sensing estimation model, find out the biochemical sensitive bands of fir, make a tentative study for the spectral characteristics of forest species, and do basic research for building up fir hyperspectral remote sensing estimation model with different age group and different time in the future.