ON-BOARD CHANGE DETECTION WITH NEURAL NETWORKS

The aim of this paper is to describe a potential on-board change detection chain by earth observation satellites (optical and SAR). The benefits of such an on-board chain are multiple : reduction of the amount of data to transmit from board to ground and autonomy of the system for example. We describe particularly one component of the chain : a detection/classification module based on the neural network, this type of algorithm being a part of a more global future fusion-based classification module. First experiments have allowed to validate the algorithm based on neural networks and first results are satisfying. The continuation consists first to enlarge the classification module with the implementation of other classification algorithms and to compare them with a more exhaustive set of data.