Context and objectives
There are currently an estimated 100.000.000 mines lying hidden throughout the world. Current humanitarian demining methods are slow and costly. Advanced methods for minefield survey are needed. The potential contribution of temporal satellite image sequences for minefield delineation has not yet been proven. Typical mine field indicators include changes in vegetation, track traces of military activity, abundant military equipment, changes of land use due to population migration, modified human activity in and around villages... These indicators can probably be identified from change detection in temporal satellite image sequences. The objective of the current project is to show that temporal satellite image sequences are a useful source of information for enhanced detection of minefield indicators. It will be shown that different methods of change detection in temporal satellite image sequences allow the identification of important mine field indicators. For that purpose, the feasibility study will focus on a Belgian test field in Leopoldsburg for which SPOT satellite imagery is available. After a study and a selection of relevant change detection methods, thematic change detection maps will be derived for the Belgian test field from a sequence of SPOT images that are selected for time frames that are critical for the mine field construction. The choice of a Belgian test field with known ground truth is known, makes its possible to verify the obtained results and to validate - more generally - the potential role of satellite image sequences for mine field delineation.
Expected scientific results
The output of the project includes apart from rectified Landsat TM, Landsat MSS and SPOT images, 8 change detection maps : 4 modulus change maps TM840903-TM920909, TM840903-TM950902, TM950902-SPOT990828, MSS7208-MSS7908, 3 absolute change maps TM840903-TM920909 r,g,b and 1 biased change map TM840903-TM92090. Different arguments are given to motivate our conclusion that the interpretation of radiance differences of local changes in the change detection maps is extremely difficult. The analysis of local structural characteristics of the changes is not possible due to the typical dimension of the minefield with respect to the resolution of the available image data. These results will be discussed with the user partner of the project, Handicap International, during a workshop at the VUB and will be made available for the PARADIS project of RMA.
|Project leader(s):||VUB - IRIS Multidimensional signal processing and communication|