The Study On Forest Classification Based On The TM Data

Take the state forest farm in Baishitougou Tomato Left County Hohhot as the object of study, study classification of the main arbor tree species by taking advantage of TM data of the growing season and non-growing season.Establishing the training samples of main arbor tree spices needed to be classified based on Mahalanobis distance and Fisher discriminate by using TM data of the growing season, extracting the gray values of TM4、3、2band combination as eigenvector, and taking the mean vector set up discriminate formulars, classify the main arbor tree species of the study area. Then combining with the inventory data of study area, testing the classification result. The result indicated that the total classification accuracy is67.73%based on Mahalanobis distance, the classification accuracy of Larix principis-rupprechtii、Pinus tabuaeformis、Betula platyphylla respectively are39.79%、41.62%、88.17%. The total classification accuracy is67.25%%based on Fisher discriminate classificatio, the classification accuracy of Larix principis-rupprechtii、Pinus tabuaeformis、Betula platyphylla respectively are62.16%、61.27%、70.98%.Extracting Pinus tabuaeformis forest by combining TM data of different phases, obtained classification accuracy is80.49%based on the relevant mask; after conducting difference computing of TM4(R)3(G)2(B) band, obtained classification accuracy is86.82%by use of maximum likelihood method; obtained classification accuracy is69.75%by use of density segmentation; obtained classification accuracy is82.99%after conducting band difference computing and density segmentation by use of TM data of the non-growing season. The result indicated that the classification after band computing is superior to the Mahalanobis distance classification and Fisher discriminant classification by combination phase characteri(?)cs of tree species in remote sensing image and its own growth characteristics...