Forest Resource Classification Research Based On SPOT5

In this research, the study area is Ma Xiao town, in Lin’an city. Combined with the SPOT5images,DEM (digital elevation models) and data of local forestry resources survey, we research theclassification of woodland and non-woodland, coniferous and broadleaf forest, as well as between thedominant tree species. We aim to find the most appropriate classification method and synthetic images.Firstly, we preprocess the SPOT5images including ortho-rectification, radiometric calibration,atmospheric correction, geometric registration, and analysis the slope and aspect of DEM; Secondly,through a combination of bands between images, we get five target images; Finally,we use maximumlikelihood and decision tree method to research classification of these images respectively.After the analysis of classification, we get the following conclusions.(1) The accuracy of theSPOT5image can be achieved good result by prior removing of the miscellaneous forest and purifyingthe forest species.(2) The atmospheric correction can affect the accuracy of the image classification.After atmospheric corrected, the accuracy of the classication is significantly higher than non-corrected.(3) The accuracy of the classification can be further improved when the DEM join the classification.(4)For the classification of woodland and non-woodland, the accuracy of using the decision tree method isalmost as same as that of maximum likelihood classification, they can reach98.03%(The kappacoefficient is0.9386); For the classification of coniferous, broadleaf and bamboo forest, the accuracy ofusing maximum likelihood classification,its classification accuracy reaching94.46%(kappa coefficientis0.8939), is higher than that of using decision tree method; For the classification accuracy of thedominant species,the accuracy of using maximum likelihood classification, its classification accuracyreaching80.59%(Kappa coefficient is0.7293), is higher than that of using decision tree method. Thiskind of classification can provide reference for future forestry resources survey.The innovation of this paper:(1) To the SPOT5images, three-tier classification can increase thecontrast of the classification, which study the potential of SPOT5image in classification of forestresources.(2) By compositing various auxiliary data with SPOT5image one by one, we can find whichauxiliary data can have the maximum impact on the classification.