Methods Of Airport Target Detection Based On Texture And Structure Analysis

With the rapid development of remote sensing, computer and network technology, acquisition of remote sensing images with high resolution and high precision has become possible. At present, the issue that how to retrieve remote sensing information that we need from a large image database quickly and accurately is a valuable research focused at home and abroad. Airport is a common target extracted from remote sensing images, the detection of which has important strategic significance.Aiming at large area of remote sensing images, an airport detection algorithm based on the geometric shape and texture characteristics is proposed in this paper. It extracts image texture feature vector according to the principle of gray level co-occurrence matrix and classifies them by the method of support vector machine(SVM) to select region-of-interest(ROI) of the airport area. On the other hand, linear feature is extracted by the improved Hough transform method to find characteristics of possible airport. Finally, the overall outline of principal runway is extracted by region growing algorithm. Experiments show that the algorithm can detect airport regions from different images correctly. The main jobs and innovations are as follows:1 Explains the concept and the background of the support vector machine, and applies it to identify the airport area. According to the principle of the airport area texture feature differing from other areas, the known field texture patterns studied by the algorithm of support vector machine are used to detect the airport area. The steps roughly include image resampling, edge detection, texture feature vectors construction, training function, classification recognition and extraction etc.2 The Hough transform is studied in this paper to extract the linear feature. The classical Hough transform is introduced, and a certain improvement for it is put forward to detect the airport runway witch has obvious characteristics of linear structure. This improvement mainly solves the questions that the classical Hough transform cannot get the line endpoint, and unable to distinguish between different segment on the straight line extension.3 In view of the poor effect of line extraction in fuzzy image by traditional algorithm, a fuzzy linear detection method is proposed to detect the inconspicuous linear feature, then this paper verifies and analyzes the effect of this algorithm through actual images. The experiments show that the method has a certain effect.