This paper introduces a new approach for SAR-based change detection over power line corridors by using relative density-ratio estimation. Unlike traditional distribution function based change detection approaches which estimate distribution probability densities of two SAR images respectively, our method directly estimates the ratio of the distribution probability densities to generate difference map. Then the difference map is clustered into changed or unchanged areas by K&I thresholding and Markov random field (MRF) based spatial smoothing. The experimental results show that this method can obtain satisfactory change detection results.
展开▼