远距离多源图像融合系统实时配准设计  

Real-time registration for long-distance multi-source image fusion system

作　　者：张宝辉[1] 张俊举[1] 苗壮 常本康[1] 钱芸生[1] 

机构地区：[1]南京理工大学电子工程与光电技术学院,江苏南京210094 [2]微光夜视技术重点实验室,陕西西安710000

出　　处：《应用光学》2013年第3期436-441,共6页Journal of Applied Optics

基　　金：国家自然科学基金(61101195);江苏省研究生创新基金项目(CX09B-097Z);微光夜视技术国防科技重点实验室开放基金(J20110506)

摘　　要：为了实现全天候多波段远距离实时图像监控,设计了具有微光、红外和可见光融合的光学前端,对多源图像进行实时配准研究。在平行光轴的基础上,通过计算不同视场图像的成像视差,计算仿射变换需要的参数,采用双线性内插算法弥补红外在成像方面与可见光图像的差别,将红外图像的变换制作成查找表存储在图像处理器DM642中,系统通过硬件查找表可以快速实现不同图像的配准变换,实现同步视频的实时配准与融合。实验表明:该设计能够准确地实现多源图像的实时配准,系统经过图像配准、图像融合和伪彩变换处理后的时间约为24.3ms,系统探测距离大于3km。In order to achieve all-weather, multi-band, long-distance and real-time image moni- toring, the optical front-end with low light level television, infrared and visible light system was designed, real-time multi-source image registration was investigated. Based on the parallel optical axis, the affine transformation parameters could be obtained by calculating the imaging disparity of the different field, the bilinear interpolation algorithm was used to make up the difference between infrared image and visible image in imaging, the infrared image transforma- tion parameters were made into a look-up table stored in the image processor DM642, the sys- tem could quickly register different images through the look-up table and achieve real-time reg istration and synchronous fusion video. Experimental results indicate that this design can reg- ister multi-source images in real-time, the total processing time of image registration, image fusion and pseudo-color transformation is about 24.3 ms, the detection range is greater than 3 km.

关 键 词：图像配准 平行光轴 双目成像 仿射变换 图像融合 

分 类 号：TN202[电子电信—物理电子学] TP391.4[自动化与计算机技术—计算机应用技术]