A SEMI-AUTOMATED METHODOLOGY FOR BURNED AREA MAPPING USING HJ-CCD DATA IN CHINA

摘要

This study focused on the development of arnsemi-automated methodology for burned area mappingrnusing HJ-CCD image. The method can be divided intorntwo main phases. First, to minimize commission errors,rna semi-adaptable threshold has been used to extractrnmore severely burned pixels as core pixels based onrnspectral reflectance and vegetation indices (VI). Then,rnto minimize omission errors and to enlarge the severelyrnburned patches previously, a logistic regression modelrnhas been made built on the information ofrnmulti-temporal vegetation indices and spectral channels.rnThe core pixels extracted in the first phase have beenrnprocessed to apply a region grow algorithm constrainedrnby the result of canny edge detection algorithm, and thernburned area can be shaped. The results showed thisrnapproach using HJ-CCD sensor produced promisingrnresults for study areas and could be suitable for burnedrnarea mapping.