Abstract:

In order to improve the efficiency of bamboo resource investigation, and supply the reference for reasonably developing the resource and planning scientifically, the author took the Taoyuan County of Hunan Province as the research object with medium resolution Landsat TM remote sensing image and 2 class survey resources distribution maps of Taoyuan County for the data source. Using of ENVI 4.5 on Landsat TM by image preprocessing, using unsupervised classification, maximum likelihood classification, Mahalanobis distance classification, minimum distance classification 4 classification methods of bamboo information extraction, and its accuracy was evaluated. The results showed that: unsupervised classification, maximum likelihood classification, minimum distance, Mahalanobis distance overall classification accuracy were 60.47%, 92.15%, 71.70%, 82.81%, respectively, Kappa coefficients were 0.4263, 0.8890, 0.6085, 0.7595. Supervised classification accuracy was higher than the unsupervised classification, and the maximum likelihood classification overall accuracy as well as the user accuracy. Kappa coefficient was higher than other 3 kinds of classification accuracy, at the same time, other types of vegetation classification accuracy could be satisfied with the results, so the maximum likelihood classification was the ideal method of bamboo information extraction.