基于异构多核构架的红外与可见光实时融合系统

A Real-time Infrared and Visible Image Fusion System Based on Heterogeneous Multicore Processor

描述了一个自主研制的基于异构多核构架的红外与可见光图像实时融合传输系统的设计与实现方案。本系统是具有异构多核并行计算机体系结构的嵌入式高速实时图像融合处理系统,选择基于ARM与DSP组合异构双核处理器TMS320DM6467T作为中心处理单元,充分利用ARM端的传输控制功能与DSP端的超强计算能力相结合的特点,发挥2种处理器构架的性能优势。提出并实现基于"灰度世界"算法的红外图像增强方式,同时使用拉普拉斯金字塔变换对红外与可见光图像进行实时融合。实验结果表明,采用该异构多核构架的图像实时融合传输系统能够良好地解决多源图像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,提高了多传感器实时图像融合处理与传输系统的处理效率和性能。

The design and implementation of a real-time infrared and visible image fusion and transmission system were presented based on heterogeneous multicore processor. It is a high speed and real-time image fusion embedded system which has architecture of heterogeneous multicore parallel. Heterogeneous multicore processor TMS320DM6467 T is used as CPU for image fusion processing. This system combines the control capability of ARM and computation power of DSP,which meets the requirement of real-time processing. This system also implements image enhancement with Grey World algorithm and image fusion with Laplacian pyramid transform. The system fairly balances heavy computation burden and real-time requirement to some extent. The experiments indicated that the image fusion system works steadily and reliably. The heterogeneous computational architecture significantly improves the efficiency and performance in realtime fusion and data transmission.