Research On Airport Detection In Remote Sensing Images

This paper is concerned with airport detection. In order to detect airport from remote sensing images, this paper dives into such techniques as feature extraction of airport runway, extraction of straight lines, region growing to obtain the architecture of airport runways. The experiments about these techniques have been performed, and the results show that the detection of airport is quite promising. The works completed and accomplishments achieved in this paper are given as follows.1. Three improved methods of straight-line extraction, Hough transform, phase grouping and line detection algorithm based on chain code, have been proposed.In improved Hough transform, the information of gradient phase of straight lines is used to limit the parameterθinto a small range, and only the pixels whose gradient phase locates within this range are used in transform. So the amount of computation has been reduced.In improved method of phase grouping, 8 direction masks are used to detect the direction of the image intensity gradient at each pixel. The direction of the mask which gets the biggest response at a pixel is regarded as the direction of the edge at the pixel, and the perpendicular direction is the direction of gradient. This method was tested and the result shows that the edge line-support regions are more complete. Then, the straight line corresponding to each edge line-support region is identified using weighted least square method.The line detection algorithm based on chain code is often affected by chain code tracing way. In this paper, the chain code tracing way has been improved. Then a fast and practical algorithm is proposed to detect straight lines in chain code. This algorithm can overcome the problem that directly using Freeman criteria is impractical.2. Straight-line segments are grouped into longer straight-line segments by the principle of collinearity in perceptual organization. In addition, a method of straight-line segments grouping in matching parallel lines is proposed. By this method, the parallel lines corresponding to runway in airport images can be extracted more completely.3. When the parallel lines corresponding to runway has been extracted, extraction of"L"junctions is performed at the two ends of the parallel lines. Then the region of runway can be identified.4. The runways of an airport often have the same intensity in an image. Based on this fact, region-growing method can be used to obtain the whole architecture of runways in the airport. In this method, the pixels in the identified region of runway are regarded as seed points.