Automatic Extraction of Burned Area Based on QUEST Decision Tree

Kunming was taken as the study area to study the feasibility of forest burned area extraction in large-scale range.The vegetation,water,burned site index and texture features were extracted from Landsat TM images in 2005 and 2006,then the best characteristics bands were chosen,which adapted for the large-scale range;Decision trees were determined using the feature-fused image,and fire burned information were automatically extracted in large-scale range.The research showed that the classification accuracy of the QUEST decision tree reached 84.5%,which was better than CRUISE 2D algorithm and maximum likelihood classification;The decision tree classification based on the QUEST algorithm was used to extract 97% of 3·29" burned area,but it was not ideal on large-scale.The smaller patches need to be manually removed;Finally,the "3·02" burned area in 2015 verified that the QUEST decision tree method had certain universality.This model had important application value for the statistics of forest burned areas in large scale.