Satellite On-Board Change Detection via Auto-Associative Neural Networks

The increase in remote sensing satellite imagery with high spatial and temporal resolutions has enabled the development of a wide variety applications for Earth observation monitoring. At same time, it requires new techniques that are able to manage amount data stored transmitted ground. Advanced on-board processing answer this problem, offering possibility select only interest specific application or extract information from data. However, computational resources exist limited compared ground segment availability. Alternatively, such as change detection, images containing changes useful worth being sent In paper, we propose detection scheme could be run on-board. It relies on feature-based representation acquired which is obtained by means an auto-associative neural network (AANN). Once AANN trained, dissimilarity between two evaluated terms extracted features. This can subsequently turned into result. study, presents one first yielded encouraging results set Sentinel-2 images, even light comparison benchmark technique.