An autologistic regression model for increasing the accuracy of burned surface mapping using Landsat Thematic Mapper data

摘要: An autologistic regression model, which takes into account neighbouring associations, was developed and applied for burned land mapping using Landsat-5 Thematic Mapper data. The integration of the autocovariate component (estimated a moving window 3 @ pixels) ordinary logistic model increased significantly overall accuracy from 88.18% to 92.44%. In contrast, derived with application post-classification majority filters, follow same principles, were not different that regression.