Regular Shapes Detection in Satellite Images

摘要

This paper presents a novel and fast method to distinguish regular and irregular regions in satellite and aerial images. This can be employed in further processing such as detecting man-made or artificial structures, disaster response, road sign detection and etc. In order to achieve this goal, we offer a suitable definition of regular shapes. Using this definition, we propose an appropriate technique to evaluate regularity .In our approach, discrete Fourier transform plays the main role for searching the regularity. We suggest a criterion; called RCNR, which indicates to some extent that the shape is regular. The computed RCNR value should be compared with a pre-assigned threshold value for making decision. Our suggested method has a good robustness against rotation and scale variation and could be easily implemented. Moreover, our approach is independent of the shape and has a low processing time. The experimental results show the capability and feasibility of the proposed technique in the classification of regions into regular and irregular shapes in satellite images.