The Study Of Artificial Target Detection Based On Fractal Feature

Artificial target detection is a very crucial research subject in image processing field and computer vision, which is the basis of target tracking and recognition. It has been widely used in intelligent traffic, detection of the city road and detection of the military target. Therefore, it is of great significance for the detection of artificial target. The fractal feature is excellent in distinguishing artificial target from background. Thus detection of the artificial target at high resolution visible spectral remote sensing image is primary research subject in this paper.The theory of fractal is surveyed systematically and six classical fractal-dimension algorithms have been implemented in this paper. Through the experimental analysis and comparison, the counting box and blanket covering which is of great performance has been chosen for the continual experiments.Based on the two algorithms, four different fractal features has been distilled, and been compared through the detection experiments of artificial target. A new fractal feature based on analysis and comparison of the four fractal feature has been proposed in this paper. This feature enhances the difference between artificial target and background without increasing the computational complexity.Classical double parameter constant false alarm rate algorithm has been researched and a novel rapid constant false alarm rate algorithm which can reduce the computational complexity effectively has been proposed in this paper.An novel rapid artificial target detection algorithm based on combinative fractal features is proposed in this paper. In this method, SUSAN detection has been introduced in dealing with the image firstly, then the combinative fractal features have been calculated inside and in the neighborhood of detection areas. Finally, utilizing rapid constant false alarm rate detection for the combinative fractal features, the detection results are obtained. It is been proved that the algorithm is effectively and rapidly in detecting the artificial target, with high probability of right detection and low false alarm.