We need to identify landslide prone areas with SPOT5 and make a robust rule set in Ecognition software. Do you have any ideas on how to do this better?
I suppose, you may use some parameter, which describes the vegetation, for example NDVI, because tree roots strengthen the soil layer. Also you should use topography maps, from which you detect the zones with large slope of land surface.
If using NDVI, you should take into account the vegetation cycles, so I think, that the method would be usable only in summer.
Also, it does not give information about geological structure of the territory. Maybe to estimate the danger of sliding of large masses of land, you will have more useful information from geological structure and topography maps of region.
I think you need a DTM, geological/geotecnical information and use a drone to collect high resolution images and construct a rule over all this data on Ecognition. If you are not a geotecnical expert you should get some expertise on slope stability to construct the rule.
Some work has already been done and published on characterisation and detection of landslides from satellite image and DEM using object-based methods. Pl. go thorugh the link http://www.itc.nl/Pub/OOA-group/Data_and_rule_sets.html for rule set developed in eCognition to detect landslides and send the feedback!