林地背景下伪装目标偏振成像检测算法

Algorithm for camouflage target detection based on polarization imaging in woodland background

综合利用了偏振特征和偏振融合图像的分形特征，提出了一种林地背景下伪装目标偏振成像检测算法。首先通过模糊聚类的方法分割出偏振图像中目标的潜在区域，通过形态学方法去除微小干扰；然后在潜在区域提取分形特征，利用伪装目标与背景分形维数的差异排除背景奇异区域的干扰；最后对记录的目标区域进行判决，得到真实目标的位置和形状特征，从而检测出伪装目标。实验结果表明，伪装目标区域的分形维数在2．0～2．3之间，而林地区域的分形维数较大，变化幅度小，且与目标的分形维数差值在0．2左右。利用伪装目标与林地背景分形维数的差异可以有效地进行伪装目标检测。

Synthetically utilizing polarization characteristics and fractal characteristics, an algorithm based on polarization imaging was proposed to detect camouflaged target in woodland background. Firstly, fuzzy clustering method was used to segment the regions containing potential target in polarization image, and the morphological method was used to remove the small interference. The information of these regions was recorded. Secondly, fractal features were exacted from these regions. The fractal differences were used to select out the background. Finally, regions with different fractal features were judged. The position and shape of the targets were determined. The experimental results of real scene show that the fractal dimension of camouflage target is between 2.0 and 2.3, the fractal dimension of woodland background is higher while its change is little and the difference between each other is about 0.2. It proves that this method is validity and reliability of the scheme which based on the fractal dimension difference between the camouflage target and woodland background.