Transferability of OBIA rulesets for IDP camp analysis in DARFUR

The analysis of refugee and IDP (internally displaced persons) camps from VHSR (very high spatial resolution) satellite imagery can assist humanitarian relief organisations by providing population estimations and camp structure analysis based on automated dwelling extraction. Since smooth transferability of rulesets in a fine scale and high complexity environment is limited, we present an approach of developing and transferring master rulesets. The master ruleset aims to automate OBIA (object-based image analysis) tasks which can be seen as more or less independent from different sensors and geographic areas, such as adapted segmentations based on edge densities to delineate man-made structures or classifications based on shape and context. Only some steps within the classification process which are based on hard-coded values (size constraints, spectral thresholds) have to be adapted. This approach reduces the time required to analyse different IDP camps or the same camp at different times. In previous work, the master ruleset has been developed for the IDP camp Zam Zam in Darfur, Sudan, using QuickBird imagery. In the study presented, the transferability of object-based image analysis rulesets was tested in a demonstrator exercise according to a request posed by the World Food Program (WFP). The aim was to extract dwelling structures and variations in dwelling densities in three IDP camps in West Darfur for estimating population figures using GeoEye-1 imagery. Results were validated using independent visual interpretation, showing high agreement for all dwelling density classes and absolute numbers of extracted dwellings (15,349 automated dwelling structures versus 14,261 visually extracted dwelling structures) for the IDP camp Um Dukhun. In the other two camps, however, the areas of higher dwelling density were apparently underestimated by the automated approach. * Corresponding author.