Skip to Main Content
An effective method is presented to rapidly detect circular-object clusters in large remote sensing images. The method uses a sliding window to cut the original image into several sub-images to accelerate the detecting speed, in each of which circles are detected by adopting a circle detection algorithm based on shape parameters. In terms of their number and their spatial distribution regulation, these circles are verified to obtain the real circluar objects and to detect the circular-object clusters by a region-growing-based clustering method. Test results from a series of experiments on large natural images (up to 10,000times10,000) indicate that average correct recognition rates of about 85% and the average running time of about 8 s can be achieved by using the proposed method.