From a change detection image to an operational alert system with Sentinel-1 time series

摘要

Full automatic change detection is a great challenge for many institutions, and many teams are involved in it: teams in charge of monitoring designated sites, teams in charge of imagery acquisition, and teams producing maps. The large availability and repetitiveness of free Sentinel-1 imagery appears like an opportunity to monitor change detection on designated sites or on larger areas. This paper presents a change detection workflow built on SNAP and Sentinel-1 time series answering operational needs: automation, scalability and usable information. It also details the tests drove on various sites and changing contexts (refugee camps, airport infrastructures, earth wall, etc.). The robustness of the process is discussed, the way to move from site monitoring to a whole regional area change detection and how external data and post-processing on a change detection image can strengthen the process in order to provide an operational full automatic alert service.