基于可见光光谱图像的红外多光谱图像仿真生成  

Infrared multispectral image simulation based on spectral images in visible bands

作　　者：许洪[1] 王向军[1] 刘峰[1] 张召才[1] 

机构地区：[1]天津大学精密仪器与光电子工程学院精密测试技术与仪器国家重点实验室,天津300072

出　　处：《红外与激光工程》2009年第2期200-204,共5页Infrared and Laser Engineering

基　　金：国防重点预研基金项目

摘　　要：阐述了红外多光谱图像仿真技术的意义和原理,研究了一种红外多光谱图像的仿真生成方法。提出了一种基于可见光/近红外波段多光谱、超光谱图像数据的地面场景建模方法,以及无监督分类方法和有监督分类方法相结合的地物像元分类、匹配、标记的策略,可以高效地解决像元地物自动匹配标记的问题。利用RGB彩色图像验证了这一方法,在将图像分割后为每类像元赋予相应的红外发射率数值,生成了4个红外波段的多光谱仿真图像,验证了该方法的可行性,指出了多光谱、超光谱图像数据在仿真应用中的各自特点。从仿真结果可以看出:不同波段图像中目标和背景之间呈现不同的特征。该方法可以生成空间形貌和辐射特征接近真实环境的红外多光谱仿真图像,对长波红外波段的多光谱成像探测仪器的研制和目标、背景光谱特征分析与探测算法的研究具有一定意义。A kind of infrared multispectral image simulation method was researched after introducing the significance and principle of infrared multispectral image simulation. An infrared multispectral ground scene modeling technology was proposed based on multispectral or hyperspectral images data in VIS/NIR band. A combined strategy of unsupervised and supervised classification was put forward to efficiently solve the auto- matching and marking of pixels. The method was verified by RGB images. Simulational images in four LWIR bands were generated after pixels segmentation and infrared emissivity assignment. The respective characteristics of multispectral and hyperspectral image data in pixels classification and matching were pointed out. Simulational results show the difference between targets and backgrounds. Using this method, infrared multispectral simulation images can be generated with geometrical topography and radiative characteristics highly similar to real natural environments, which are valuable to development of LWIR multispectral imaging spectrometers, the spectral signature analysis of targets and backgrounds as well as the development of detection algorithms.

关 键 词：红外多光谱图像仿真 多光谱图像 超光谱图像 无监督分类 有监督分类 

分 类 号：TN21[电子电信—物理电子学] TP7[自动化与计算机技术—检测技术与自动化装置]