针对传统的红外、可见光等单波长侦察手段在复杂伪装背景下难以准确有效的提取目标的问题, 提出了一种基于超光谱图像的快速反伪装识别技术。首先, 对实时采集的多谱段图像进行配准并进行图像融合获得超光谱图像; 然后, 对融合后的图像进行DDE增强, 凸显图像细节信息; 最后, 对细节增强后的图像进行阈值分割和连通域提取, 获得感兴趣的目标。实验结果表明, 该方法实时性好、消耗CPU缓存资源少, 能将单一波段难以检测的复杂伪装下的房屋营地等目标可以快速清晰分辨出来。

A quick anti-camouflage recognition technology based on hyper-spectral image is proposed for resolving the problem that the target is not easy to be extracted accurately and effectively with traditional infrared and visible light single wavelength reconnaissance methods under complex camouflage background. At first, the collected data of multi-spectral image in real time is registered and fused to obtain hyper-spectral images. And then, the fused images are enhanced by digital detail enhancement (DDE) technology and image details are effectively highlighted. At last, the methods of threshold segmenting and connectivity area abstracting of the enhanced image are used to obtain the targets of interest. Experimental results show that the method has better real time and less consumption of central processing unit (CPU) resources. It can quickly and clearly distinguish the targets such as buildings and campus under complex camouflage background, which is difficult to detect for single wavelength methods.