SAR scene potential for the monitoring of anthropic activities (deforestation, development) through Markovian classification techniques and tree-structured wavelets based classification algorithms

The general objective of the study consists of monitoring the temporal evolution of small parcels of deforestation (abattis in French) and natural non-forest areas in French Guiana. A comparison is done between the results obtained through photo-interpretation and those obtained by segmentation and classification techniques adapted for radar images (Markovian classification and tree-structured wavelets based classification algorithms). In the first study site, the results of classifications constitute a data source to update the cartography of the ground occupation. Concerning the impact studies, the techniques of linearization by Markov fields contribute to a better estimate of the dike effect tied to the causeway of the road route.