基于局部自相关性的特异性光谱谱段选取算法

Algorithm for Selecting Specific Spectrum Based on Local Self-correlation Property

高光谱成像探测技术已经广泛应用于伪装物探测领域,但丰富的光谱细节意味着巨大的数据量,这导致数据存储和计算处理过程时间很长,难以满足战场环境下伪装物实时探测识别的要求。传统的光谱信息压缩和快速选取算法大多需要目标的特征光谱,而某些战场中伪装目标的光谱通常难以获得,这降低了传统算法的可用性。针对战场环境伪装物的视觉伪装效果,提出了基于目标与背景间局部自相关系数对疑似伪装目标进行检测。算法通过选取最佳特异性光谱,实现无光谱先验信息的伪装物识别,提高了高光谱成像探测目标检测效率。实验分析表明,该算法可以有效提取三个特异性光谱,实现伪装物目标识别与提取过程。

Hyper-spectral imaging technology has been widely applied in camouflage target detection. However, plenty of spectral details mean a very large quantity of data, and it leads to long time for data saving and calculating, which is difficult to meet the requirements of camouflage target detection and recognition in real time in battlefield. Most traditional algorithms for spectral information compression and quick selection need the characteristic spectrum of targets, while it is always difficult to obtain in advance, thus the practicability of those algorithms is reduced. According to the visual effects of camouflage targets in battlefield, an algorithm detecting camouflage targets based on the local self-correlation coefficient between the targets and background is proposed. In the proposed algorithm, the suspected camouflage targets are recognized without its priori information by selecting the best specific spectrum, and the detection efficiency of a hyper-spectral imaging system is improved. Experimental results show that the proposed algorithm can effectively extract three specific spectrums to realize the camouflage targets identification and extraction processes.