Abstract
A supervised method for change detection from remote sensing images based on cross- and auto-correlation is presented. The work is presented as a generic approach to image change detection whereas forest fire bun scars detection just being the specific test used for validation purpose. Using only spectral information of the images for detecting change ignores the within and between image correlation and is not capable of representing change in temporal and spatial domains. The proposed method is designed to improve change detection accuracy by combining spectral information with both local cross- and auto-correlation. A novel way of a simultaneous incorporation of these associations by means of cross- and auto-correlation is put forward and tested here under the logistic regression framework. The proposed method showed the increment of 9% in Kappa accuracy as compared to simple logistic regression ran over spectral information alone. It demonstrated that the use of cross- and auto-correlation leads to substantial increase in burned area detection accuracy.
