A new method for airport remote sensing image detection based on visual saliency and spatial pyramid feature

We use the improved line segment detector algorithm to judge preliminarily whether the sliding window contains an airport or not. Then we use the spatial pyramid feature expression method based on visual saliency to extract the sparse coding of each image patch in the sliding window that may contain the airport. The global feature vector that characterizes the sliding window is formed by using the visual saliency-based feature extraction strategy and then classified and judged by the support vector machine to distinguish the airport image from background image, thus obtaining the confidence values of the sliding window that contains the airport. Finally we use the non-maximum suppression method to detect the airport image. The simulation results, given in Table 1, show preliminarily that our airport detection method can robustly express the sliding window and effectively detect the airport image, has a higher detection rate and smaller false alarm rate the other airport detection methods.