NMI-Based Small Target Detecting Method

For features of synchronous orbit target image, this paper proposes the small target detection method that based on Normalized moment of inertia feature. Firstly, this method uses gray mathematical morphological filter to pre-process the image, improves the target signal noise ratio without affecting the target shape, and then uses constant false alarm rate (CFAR) to choose segmentation threshold for the image binarization to further remove the noise interference. Then location information of each pixel are obtained by binary tags, by using the candidate target location information of each pixel to calculate target relative to the centroid of the normalized moment of inertia. Finally, the specific target is detected by appropriate threshold based on normalized moment of inertia for required target type. Algorithm is simple and fast. And by using real film images for experiment, results show that the proposed NMI-based small target detection method is simple and effective.