一种基于海量信息处理器的辨云相机系统设计  

A Cloud-Distinguishing Camera System Design Based on Mass Information Processor

作　　者：马瑶 郭冯凤 王莉[1] 刘曦[1] 肖杰 MA Yao;GUO Fengfeng;WANG Li;LIU Xi;XIAO Jie(Xi'an Institute of Microelectronics Technology,Xi'an 710000,China)

机构地区：[1]西安微电子技术研究所,陕西西安710000

出　　处：《西北工业大学学报》2019年第S01期15-19,共5页Journal of Northwestern Polytechnical University

摘　　要：针对星上数据实时处理系统设计发展需求,以可见光遥感卫星载荷实时云量覆盖估算应用需求为背景,以七七一所所产抗辐照海量信息处理器与CMOS可见光传感器分别为处理核心和成像单元设计了一款星上实时辨云相机系统。同时,设计了一种面向运算资源受限的空间多核处理器的在轨云判方法,采用神经网络分类器算法实现,在此基础上,通过并行与流水相结合的计算架构,结合海量信息处理器的核间共享内存机制,完成算法的处理器部署,实现了系统云判功能,为星上数据实时处理系统设计进行积极的技术探索。Aiming to satisfy the requirement of on-board data real-time processing system,an on-board real-time cloud-distinguishing camera system which consists of the mass information processor and CMOS visible light sensor is designed in this paper.The mass information processor produced by Xi'an institute of microelectronics technology performs as the core unit and the CMOS visible light sensor acts as imaging unit.Meanwhile,an in-orbit cloud-judging method for multi-core processor with limited computing resources is proposed in this paper.This method is realized by neural network classifier algorithm.Besides that,it completes the processor deployment and cloud-judging function by adopting the parallel and flow computing architecture and the mechanism of sharing memory between cores of mass information processor.The design of this method plays an important role in the active technical exploration for on-board data real-time processing system.

关 键 词：辨云 多核处理器 神经网络 共享内存机制 

分 类 号：TP752.1[自动化与计算机技术—检测技术与自动化装置]