The Research On The Response Of Bamboo Information And Etraction Technology Based On HJ-1A Hyperspectral Data

China has abundant bamboo resources,extracting and monitoring dynamically bamboo information quickly and efficiently will play an important guiding role in the development and construction of bamboo industry.Remote sensing technology provides for the monitoring and investigation of bamboo resources in an efficient and accurate means.In southern China,however,which has lots of hills and mountains,as well as influenced by a subtropical climate,understory vegetation growing well.So,it will exist the phenomenon of synonyms spectrum,with spectral foreign body when we identify features replying on multi-spectral remote sensing.As the material structure and growth characteristics of the vegetation are different,therefore,they will show different spectral absorption and reflection characteristics.Hyperspectral its fine spectral resolution can accurately detect the subtle spectral features,which have diagnostic significance for species identification,and it will greatly improve the recognition accuracy of forest tree species,which provides useful information for the identification of the bamboo resources on remote sensing images.In this paper,take Yong'an City,Fujian as the research area,use environmental satellite hyperspectral remote sensing data(HJ-1A HSI),with the goal of bamboo information extraction,carried out bamboo spectral response analysis on HJ-1A HSI image data,the choice of characteristic bands depending on bamboo information on HJ-1A HSI image data,the building of effective characteristic quantity that can help identify characteristics of bamboo,and determination of the best classification of bamnoo information extraction,the main conclusions are as follows:(1)Get the spectral values of bamboo,pine,fir,hardwood,economic forest on the HJ-1A HSI image,in terms of spectral correlation coefficient,spectral angle cosine,and spectral information divergence to analysis spectral similarity of these forest types on the HJ-1A HSI image.The results showed that the spectral difference among bamboo,pine,hardwood is small,and between bamboo with economic forest has quite different spectral.By mathematical transformation of the original spectral curve to analyze the spectral differences between bamboo with the other four kinds of forest types,obtained at a wavelength of 759 nm?770 nm,the 917 nm?934 nm wavelength range of these two features,the various forest types show the most significant spectral differences,bamboo has the greatest spectral difference with other forest types near the 464 nm,538 nm,565 nm,925 nm.(2)In order to choice the characteristic bands,using six methods spectral differential method:a first-order and second order spectral differentiaion method,remove the envelope method,spectral mixture distance,grouping band index method,and adaptive band index selection method.Using maximum likelihood to extract the information of bamboo and other forest types,and determine the best combination of characteristic bands from the six groups of band combination depending on the highest accuracy of bamboo information extraction.The results show that the band combination selected by the method of the first order spectral differentiaion method has a higher recognition accuracy,the recognition accuracy of bamboo reach at 67.99%.That is,the best bands of bamboo information extraction on HJ-1A HSI image are1,3,5,6,8,9,10,11,12,17,20,22,37,49,76,77,87,88,90,107,108,109,110,111,112,113,a total of 26 bands.(3)Introduction of principal components,texture features,topography factor of three characteristic quantities,respectively combined with the 26 bands determined by a first-order spectral differentiaion method,using the maximum likelihood method to extract the information of bamboo and other four forest types.By contrasting classification accuracy,we can determine which auxiliary variables to take in.The results showed that the first principal components,altitude andelevation gradient superimposed can effectively improve the recognition accuracy of bamboo information,compared with the accuracy,the recognition accuracy of bamboo increased 0.1%,0.66%,2.54%respectively,and texture features did not improve recognition accuracy,and the study found that adding a second and third principal components will reduce the principal components of bamboo information extraction accuracy.(4)Take the 26 bands and complex characteristics(the first principal components,elevation and slope)overlay,using the maximum likelihood method for classification,information extraction finally gets an overall accuracy of 69.22%,Kappa coefficient is 0.6146,bamboo extraction accuracy is 71.1 1%.(5)Using the maximum likelihood method,spectral angle matching,classification tree to extract bamboo and other four forest types.The results show that the maximum likelihood method of image classification gets a accuracy of 69.22%of the total,the user accuracy of bamboo is 71.11%;spectral angle matching the total image classification accuracy of 54.39%,users accuracy of bamboo is 50.83%;decision tree classification method of image classification accuracy of 64.16%of the total,the user accuracy of bamboo is 65.45%.So,take the maximum likelihood method as the best classification methods for bamboo information extractation on the HJ-1A HSI image data.