An CFAR detection algorithm based on local fractal dimension

Target detection is an very important research subject in image processing field and computer vision . Aim at the problem of long time detection and higher false alarm during detection artificial target under nature background in optical remote sensing image, A CFAR algorithm based on local fractal dimension is proposed in this paper．The algorithm firstly introduce rescaled range analysis ,by changing the local glide image into the sequences of one dimension, and calculating its range and deviation, produce local fractal dimension that reflects local texture feature of the image，and construct fractal dimension image of the image based on the fractal dimension. Secondly, a fast constant false alarm rate(CFAR) detection algorithm is used for fractal dimension image to ensure whether the center of the glide window pixel is the target one. Finally, a region of interest target is obtained by clustered the target pixel based on detection result. A lot of experiment has down in different region optical remote sensing images, using the algorithm proposed by this paper, and lots of good result has been gained. The experimental result prove that the algorithm can detect the artificial target under nature background effectively and rapidly in optical remote sensing image．Compared with the traditional artificial target detection algorithm, the presented method can reduce the computational complexity effectively, lessen the probability of false detection successfully.