Different fusion strategies to detect geographical objects by active contours in multitemporal SAR images

When a geophysicist has to make a visual interpretation in multitemporal SAR images, it may be long and repetitive. To avoid this, an automatic object detection using multitemporal active contours is proposed in this paper. The information brought by the different images can be fused at different levels: either at the data level, the feature level or a level close to the decision level. On two data-sets, both located in French Guyana, two different strategies will be tested depending on the knowledge of the object, whether it is temporally stable or moving.