Airport Detection in SAR Image Based on Perceptual Organization

Being one of the key transportation targets, airport detection is of great importance in military and civil applications. In this paper we propose a new method based on perceptual organization for airport detection in large SAR image. Since the runways are the most obvious characteristic of the airport, we first design belief functions for the runway features, which include collinearity, proximity, width and texture similarity. Then we propose to use the DS-evidence theory for the fusion of all belief functions, which can serve as a criterion to decide whether the two region units can be grouped. The candidate airport can be found by grouped region growing and detected by airport knowledge. Experimental results showed the effectiveness of the proposed approach for airport detection in large complex SAR image.