Airport recognition based on contextual information

An approach to incorporate contextual information into airport recognition was presented. It was helpful to reduce the ambiguity of objects in the images by utilizing other information referred to as context, which in our case was the identity of the accompanying objects. Airports were typical multi-component class of objects with obvious orientation. The original image was decomposed into multiple spatial frequences and orientation selective channels by Gabor filters. Each channel corresponded to one feature image. Radon transform was adopted to extract line features from these feature images. According to the knowledge about the structural information of airports, a series of rules to determine the airport in the image were used.