多源数据决策技术在红外目标检测中的应用  

Target Detection of Infrared Images Based on Multi-source Data Decision-making

作　　者：林晓敏[1] 张涌[1] 

机构地区：[1]中国科学院上海技术物理研究所,上海200083

出　　处：《半导体光电》2009年第2期308-312,共5页Semiconductor Optoelectronics

摘　　要：针对各个滤波算法的适用范围不同,仅仅依靠单一的滤波算法不能适应红外系统的实际需求,提出了复杂背景下红外图像弱小目标的多算法并行处理硬件结构,该结构采用了基于自适应"投票表决"的多源数据决策技术,能根据外场的实际情况自适应地将各种滤波算法的优势集合起来,可以极大地改善滤波效果,提高点目标的检测概率,为后续处理降低数据量及运算量。针对硬件仿真阶段注入原始红外图像数据流困难,无法实时对比图像处理前/后效果的问题,基于Wishbone-PCI桥核设计并实现了以FPGA为主处理器、主机为从处理器的红外图像处理仿真平台。实验表明,多源数据决策技术很好地满足了系统的实际需求,多算法并行处理硬件结构运行稳健有效,硬件仿真平台可靠稳定,整个系统有很强的工程实用价值。Aiming at the problem that single filtering algorithm can not fit the requirement of infrared warning system, a multi-algorithm parallel processing hardware architecture is introduced for detecting infrared point targets under complex background. This architecture adopts multi-source data decision-making based on self-adaptively voting, which self-adaptively fuses advantage of kinds of filtering algorithm depending on the actual condition, the filtering effect can be greatly improved, and the operation complexity can be greatly reduced. An infrared image processing emulation platform based on Wishbone-PCI Bridge Core is designed and implemented. The emulation platform solves the difficulty in injecting raw infrared image and comparing the effects of images before and after treatments intuitively and real-time during hardware emulation. The emulation experiment based on real infrared image shows that, the multi-source data decision-making greatly satisfies the requirement of infrared warning system, multi-algorithm parallel processor hardware architecture is effective and robust, the emulation platform is reliable and stable, and the system has wide applications in projects.

关 键 词：多源数据决策技术 多算法并行处理 Wishbone—PCI桥核 FPGA 仿真平台 

分 类 号：TN216[电子电信—物理电子学]