Abstract:
The proposed project deals with the automatic detection of vehicles, particularly military
vehicles in high resolution aerial imagery. The extraction relies upon local features of
vehicles. To model a vehicle on local level, a model representation is used that describes
the prominent geometric features of vehicles. The model is adaptive because, during
extraction, the expected saliencies of various edge features are automatically adjusted
depending on viewing angle, vehicle color measured from the image, and current
illumination direction. The extraction is carried out by matching this model ”top-down” to
the image and evaluating the support found in the image Hence, training data samples of
vehicles are first clustered and statistical parameters corresponding to each cluster are
obtained. Vehicles are detected by searching the test image for patches of vehicles at all
points in the image and across different scales. Applying this technique to the military
vehicles particularly fighting vehicles presents peculiar problems of its own as they differ
in geometric and statistical representation from that of the soft vehicles. The project is
aimed at facilitating automatic aerial imagery analysis, which is a very tedious job if done
manually, simultaneously maximizing the accuracy and performance.
