The objective of the proposed project is to develop a robust and sufficiently accurate method for the segmentation, classification and change detection of remotely sensed data for land-cover mapping. The project will be carried out in cooperation with the host institute and with an industry that has strong links with end-users of remote-sensing data. Special emphasis will be placed on two aspects:
1. a methodology to assess the quality of segmentation, classification and change-detection methods;
2. the systematic development of the aforesaid methods which should be more cost-effective than existing methods.
Innovative aspects are:
1. explicit definition of quality measures in cooperation with industry; 2. extensive use of different types of information present in data; 3. high accuracy according to defined quality standards; 4. high levels of user-friendliness and robustness provided by the automatic estimation of processing parameters that will aid a future implementation in geographical information systems.//ORLhttp://dibe.unige.it/tmr_smits