A burned area mapping method through fire location clustering and vegetation index for Medium Resolution Spectral Imager

Extracting the burned area can evaluate the loss of economy and vegetation. Remote sensing technology can be used in the mapping of burned areas. Most previous methods extracting burned areas based on the changes of vegetation index (VI), but these changes may not completely caused by fire. This paper introduces an accurate burned area mapping method through fire location clustering and VI for FY3D Medium Resolution Spectral Imager-II(MERSI-Ⅱ) data. Firstly, select cumulative map of fire location obtained from MERSI-Ⅱ mid-infrared and far-infrared brightness temperature, the cumulative results are filtered and clustered as initial samples. Then Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is used to cluster initial samples into burn samples. Besides, we can calculate the coverage of the burned area through multiple spectral data by using the reflectance and the tempertature. We chose normalized burn ration(NBR) based on near infrared(NIR) and short wave infrared channels, and vegetation index thermal(VIT) based on NIR and long wave infrared channels. Finally, NBR and VIT are used to expand burned samples to extract burned area by region growth algorithm. We compare the proposed method with the true color map after fire and the Collection 6 MODIS burn area products (MCD64A1) for verification. The results show that the proposed method has a high correspondence with true color map. The commission errors is lower than MCD64A1, with 0.1703 and 0.4671 respectively. The kappa coefficient is significantly higher than that of MCD64A1, with the two being 0.8429 and 0.6272 respectively.