摘要
描述了一个自主研制的基于实时分布式多处理机的图像配准和融合系统的设计与实现方案。本系统是具有并行计算机体系结构的通用高速实时图像融合处理系统,选择 VxWorks 实时操作系统和VEM64x 总线的软硬件平台,采用 AD 公司新型的 TS101DSP 处理器为核心,多 DSP 处理器分布并行进行处理,完成多源图像实时高速配准和融合需要进行的大量运算,CPLD 芯片完成了采集控制以及多传感器视频同步。由于采用了基于高性能 DSP 的实时嵌入式系统和通用标准化总线结构设计, 该系统可以灵活地应用多种配准和融合算法来实现可见光和红外双通道数字图像的高速实时融合处理, 比较好地解决多尺度图像配准融合算法的大数据量计算处理与系统实时性要求之间的矛盾, 为多传感器实时图像配准融合处理系统的研制奠定了良好的技术基础.
The scheme and implementation of a Real-Time Multisensor image registration and fusion system based on ADSP-TS101DSP is described, it is a universe high speed and real-time image registration and fusion system which has architecture of parallel distribution. Selecting VxWORKS RTOS and VME64x bus as platform for multisensor image fusion processing. TS101 DSP is served as the CPU to perform registration and fusion parallel and real-time. CPLD is designed to control VMEbus interface, logic of circuit, video synchronization. Some specific problems on hardware designing for image fusion are also discussed. As a result of using high performance DSP and standard parallel structural, various processing algorithms are applied flexibly, the system can meet the requirement of real-time two channel image fusion processing from visible and infrared band. The system fairly balances heavy computation burden of multiscale based on fusion algorithms and real-time requirement to some extent. The experiments indicate that the image fusion processing system works steadily and reliably. The performance and realization of real-time image fusion processing with complex algorithms such as pyramid decomposition and reconstruction is satisfied. This job constructed the scheme to develop multichannel real-time image fusion system for the practical purpose.
出处
《红外技术》
CSCD
北大核心
2004年第5期66-71,共6页
Infrared Technology
基金
陕西省科技攻关项目(编号:2003k05-G17)
