高分辨率遥感影像共享关键技术研究与实现
发布时间：2018-11-20 05:46
【摘要】：随着航天与电子产业的发展,光学卫星数量不断增加,传感器技术不断提升,遥感影像分辨率极大提高,单景遥感影像的数据量已经越来越庞大。近年来我国的资源系列卫星产生了海量的高分辨率遥感影像,快速有效的实现这些数据共享,必将推动遥感在各行业研究与应用的发展。面向高分辨率遥感影像,传统的共享方式已经不适用,分布式计算环境下的快速共享大趋所至。本论文从数据组织管理和传输,对遥感影像共享一些关键技术进行研究,论文主要工作包括:1.在分析三种常用遥感影像重采样算法的基础上,改进了一种基于距离加权和灰度值线性拉伸的重采样算法(RADL),该算法在保证计算性能的前提下对采样精度进行优化。2.在分析四叉树切片和顺序分块切片的计算性能基础上,发展了一种与顺序分块切片相结合的构建影像金字塔的四叉树编码方法,并对该编码方法下影像切片检索展开分析。3.引入CUDA技术,对影像重采样的计算模式进行改进,并对计算过程耗时具体分析,发展了一种基于阈值选择计算模式的快速影像金字塔创建方法。4.搭建一种P2P模式下影像快速传输体系,并设计服务端集群和客户端集群任务调度的负载均衡机制,结合具体行业应用需求开发了原型系统。本文在RADL算法中,取得了PNSR提高9%的实验效果;在阈值选择计算模式的快速影像金字塔创建方法中,取得了相对传统四叉树计算模式12倍的加速效果,相对CUDA计算模式1.2倍的加速效果;在影像传输,取得了并发请求客户端数量增加,影像分发速率近似线性增加的效果,并且CPU使用率趋于平缓。上述关键技术,广泛应用于遥感数据快速共享平台,以网络发布方式向行业用户提供数据和产品服务。
[Abstract]:With the development of aerospace and electronic industry, the number of optical satellites is increasing, the sensor technology is improving, the resolution of remote sensing image is greatly improved, and the data volume of single-scene remote sensing image has become more and more large. In recent years, our country's resource series satellite has produced massive high resolution remote sensing image, the fast and effective realization of these data sharing will promote the development of remote sensing research and application in various industries. For high resolution remote sensing images, the traditional sharing method is no longer applicable, and the fast sharing trend in distributed computing environment is becoming more and more important. This paper studies some key technologies of remote sensing image sharing from data organization, management and transmission. The main work of this paper includes: 1. Based on the analysis of three common resampling algorithms for remote sensing images, this paper improves a resampling algorithm (RADL), based on distance weighting and linear stretching of gray values. The algorithm optimizes the sampling accuracy on the premise of ensuring the computational performance. 2. On the basis of analyzing the computational performance of quadtree slice and sequential block slice, a quadtree coding method combining with sequential block slice to construct image pyramid is developed, and the image slice retrieval under this coding method is analyzed. 3. This paper introduces CUDA technology to improve the computing mode of image resampling, and analyzes the time consuming in the calculation process, and develops a fast image pyramid building method based on threshold selection computing mode. 4. A fast image transmission system in P2P mode is built, and the load balancing mechanism of server cluster and client cluster task scheduling is designed, and the prototype system is developed according to the specific industry application requirements. In this paper, the experimental effect of PNSR is improved by 9% in RADL algorithm. In the fast image pyramid creation method of threshold selection computing mode, the acceleration effect is 12 times that of traditional quadtree computing mode and 1.2 times that of CUDA computing mode. In image transmission, the number of concurrent request clients increases, the image distribution rate increases linearly, and the CPU usage rate tends to be flat. These key technologies are widely used in the rapid sharing platform of remote sensing data and provide data and product services to industry users by means of network publishing.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：P237
本文编号：2343953
[Abstract]:With the development of aerospace and electronic industry, the number of optical satellites is increasing, the sensor technology is improving, the resolution of remote sensing image is greatly improved, and the data volume of single-scene remote sensing image has become more and more large. In recent years, our country's resource series satellite has produced massive high resolution remote sensing image, the fast and effective realization of these data sharing will promote the development of remote sensing research and application in various industries. For high resolution remote sensing images, the traditional sharing method is no longer applicable, and the fast sharing trend in distributed computing environment is becoming more and more important. This paper studies some key technologies of remote sensing image sharing from data organization, management and transmission. The main work of this paper includes: 1. Based on the analysis of three common resampling algorithms for remote sensing images, this paper improves a resampling algorithm (RADL), based on distance weighting and linear stretching of gray values. The algorithm optimizes the sampling accuracy on the premise of ensuring the computational performance. 2. On the basis of analyzing the computational performance of quadtree slice and sequential block slice, a quadtree coding method combining with sequential block slice to construct image pyramid is developed, and the image slice retrieval under this coding method is analyzed. 3. This paper introduces CUDA technology to improve the computing mode of image resampling, and analyzes the time consuming in the calculation process, and develops a fast image pyramid building method based on threshold selection computing mode. 4. A fast image transmission system in P2P mode is built, and the load balancing mechanism of server cluster and client cluster task scheduling is designed, and the prototype system is developed according to the specific industry application requirements. In this paper, the experimental effect of PNSR is improved by 9% in RADL algorithm. In the fast image pyramid creation method of threshold selection computing mode, the acceleration effect is 12 times that of traditional quadtree computing mode and 1.2 times that of CUDA computing mode. In image transmission, the number of concurrent request clients increases, the image distribution rate increases linearly, and the CPU usage rate tends to be flat. These key technologies are widely used in the rapid sharing platform of remote sensing data and provide data and product services to industry users by means of network publishing.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：P237
【参考文献】
相关期刊论文 前4条
1 李建勋;沈冰;姜仁贵;陈田庆;;面向影像金字塔的四叉树空间索引算法[J];计算机工程;2011年10期
2 吴良武,欧宗瑛;保持轮廓清晰光滑的灰度图像放大算法[J];计算机辅助设计与图形学学报;2002年04期
3 刘书琴;毋立芳;宫玉;刘兴胜;;图像质量评价综述[J];中国科技论文在线;2011年07期
4 李辉;;美国民用综合性对地观测系统研究[J];中国航天;2012年04期
，本文编号：2343953
本文链接：https://www.wllwen.com/kejilunwen/dizhicehuilunwen/2343953.html
