Segmentation of high resolution satellite images SPOT applied to lake detection.

High resolution satellite imagery are widely used in many applications. One of the main applications is planet monitoring for hazard assessment. The objective of our study is to develop segmentation methods of space-borne remote sensing images capable of detecting lakes over large areas in high mountainous regions. In this context three different methods are investigated on a SPOT2 image, namely: the spectral angle mapper, the irregular pyramid and the watershed with markers. Each method having its own characteristics, we propose to aggregate the three algorithms and thus improve the segmentation result. In this paper, first we present each one of these methods and the results obtained, then we present the fusion architecture and the result obtained by the comibation of the three methods.