海面弱小目标高光谱融合技术研究  

Hyperspectral data fusion technology for weak small sea surface target

作　　者：朱院院[1] 慕巍[1] 李广良[1] 王虎[1] 刘彤[1] 高泽东[1] 

机构地区：[1]西安应用光学研究所

出　　处：《应用光学》2017年第1期37-41,共5页Journal of Applied Optics

基　　金：国防基础技术预先研究计划资助

摘　　要：海面目标受海水扰动影响,难以被宽波段光电传感器探测识别。高光谱传感器可以获取海面目标和水体的光谱数据,利用目标和海水的光谱特性差异可以有效抑制海面扰动影响,提高探测识别能力。针对高光谱数据降维融合容易丢失海面弱小目标问题,研究了弱小目标光谱数据融合方法。通过相似性分类产生类矩阵,由类矩阵主成分变换的降维投影矩阵来投影变换原有光谱数据,获得降维数据矩阵。降维矩阵通过空间变换转换到RGB彩色空间,生成伪彩色融合图像。通过远距离弱小目标和水中鱼群高光谱数据,验证了改进方法的融合性能。实验结果表明:相似性分类融合方法不仅能将高维光谱数据融合成一幅伪彩色图像,还能有效避免弱小目标融合丢失问题,提高了目标探测识别能力。Aiming at solving problem of weak small target missing during dimension reduction and fusion of hyperspectral data, hyperspectral data fusion method for weak small target is studied in this paper. Class pixel matrix is produced by similarity measures, dimension reduction projection matrix is calculated by principal component transforming class pixel matrix, and dimension reduction data is obtained by dimension reduction projection matrix projecting the o- riginal spectral data. False-color result image is produced by RGB color space transforming from reduction data. Fusion performance of improved method is verified by hyperspectral data of weak small target and underwater fish group. Experiments show that similarity classification fusion method can not only get one false color image from high dimension hyperspectral data, but also avoid missing weak small targets from background, which enhances abilities of detection and identification, and has a strong application prospects.

关 键 词：海面侦察 弱小目标 光谱数据融合 降维矩阵 相似性分类 

分 类 号：TN911.73[电子电信—通信与信息系统]