| English abstract |
Terrestrial road condition awareness in tropical areas is a key requirement, specifically in rainy season. This work suggests a methodology for trafficability analysis, and its related cartographic representation, from optical and radar remote sensing and open source data. This approach is applied to Southern Chad environments. At regional scale, a dynamic classification of obstacles, soil types and vegetation is produced from Landsat and MODIS imagery, and topographical SRTM data. A spatial measure of cone index is computed for each soil class, as a function of root density and soil moisture. This result is compared with vehicle bearing capacity to produce a mobility map. At local scale, landcover features are extracted from very high resolution Quickbird imagery by a selective combination of SVM and oriented object classifications. Soil moisture is appreciated with radar signal of TerraSAR-X sensor. Road practicability is assessed depending of conditions and road surface, and daily rainfall measurements. The model structure allows a fast and an automatic update of trafficability analysis, through workflows partitioning. Thereby, it takes advantage of the whole data available on the study area potential (imagery, scientific data, field work contributions gathered from the web). |