基于相对微分计盒法的人造目标分形检测

Fractal artificial target detection basedon relative differential box-counting method

对光电目标进行自动检测与识别是导弹末端智能制导的一项关键技术。基于图像分形维计算的相对微分计盒法,提出了一套实用的自然背景中人造目标的检测方法,给出了检测窗口与尺度选择的标准与参考值。大量实验证明,分形检测方法比传统的最大对比度检测方法稳定,是一种非常有前途的方法。

Electro-optical tar get automatic detection and recognition is a key technology of intelligent guidance for a missile.Based on the relative differential box-counting method of calculating image fractal dimension,a practical approach is presented to detect an artificial target from natural scenes.The method analyzes the criterion for selecting the sizes of detection window and scale and their reference values.Vast computer experiments prove that fractal-based detection method gets stabler results than traditional maximum contrast method.Fractal-based detection method is a promising method.