Abstract

Change detection for cross-sensor remote sensing images is an important research topic with a wide range of applications in disaster treatment, environmental monitoring and so on. It is a challenging problem as images from various acquisitions have difference in the spatial and spectral domains. Change detection models need effective feature representations to estimate interesting changes, but sometimes the hand-crafted low-level features affect the detection result. In this paper, we propose a novel cross-sensor remote sensing image change detection method based on deep canonically correlated autoencoders (DCCAE). The method extracts abstract and robust features of two multi-spectral images through two autoencoders, and then project them into a common latent space, in which any change detection models can be applied. Our experimental results on real datasets demonstrate the promising performance of the proposed network compared to several existing approaches.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call