Based on profound research in the circles of binary images, two important parameters are presented: the ratio between the number of pixels on circle and its radius, and the difference between the ideal radius and the actual radius. Using these two important parameters, a new algorithm is designed for fast and efficient circular targets detection. After denoising, edge extraction and circles detection, the centre of circles and radii parameters are exported and the circles are plotted in the image which has the same size as original one but contains nothing. Large numbers of experiments show that mass circles can be detected in very large remote sensing images fast and accurately and the algorithm implemented with VC 7.0.