异构并行计算在娄底地区水土流失分析处理中的应用研究

发布时间：2019-01-08 12:20

【摘要】：水土流失是一种突发性的自然灾害，根据我国公布的全国第2次遥感调查结果的权威数据显示，中国的水土流失面积已经达到356万km~2，占我国国土总面积的37%，其中：水力侵蚀面积达到165万km~2，风力侵蚀面积达到191万km~2。对我国的国民生产以及居民的生活造成了巨大的隐患和威胁。由于水土流失具有：突发性，不确定性及灾变性等特点，，使得精确的水土流失预警成为一个世界性的科学难题。因此，如何做到准确的灾前预防和灾后展开的救援工作，是目前减轻水土流失灾害损失最有效的方法。遥感检测技术是在20世纪60年代发展起来的一种非接触的、新型检测科学技术。一般情况下我们通过遥感检测设备来探测被测实物的电波反射、散射等特性，同时结合相关实物的特性、原理进行研究的新型科技、手段。与传统的人工探测手段相比，遥感检测技术具有：观测范围大、不受时间空间限制、综合、宏观等特点。为宏观研究各种地理现象及其相互关系提供了有力的条件；遥感数据具有：信息量大、获取手段多，使得研究人员对地球的观测具有了全天候，多方位的能力；获取信息快，更新周期短，具有动态检测等特点。基于以上的目的，本研究主要的工作和内容如下：本文首先介绍了遥感检测技术应用于灾害信息提取的研究现状，并对并行计算的相关概念、发展历程以及当今主流的并行计算编程环境、模式进行了简单的概述。在此基础之上，分析了并行计算技术在遥感图像处理中的应用的现状，介绍了基于Berkerley NOW的MPI和OpenMP混合编程模型的搭建过程，以遥感图像检测通常使用的相关系数法和比值法这两种常用的变化检测方法作为对比信息的提取方法，在本研究中所构建的机群系统上实现了基于主从模式的并行模型，最后采用湖南省娄底地区大石山附近的遥感图像作为实验采样区进行对比实验，并对并行计算的效果进行了讨论。
[Abstract]:Soil erosion is a sudden natural disaster. According to the authoritative data of the second national remote sensing survey published by our country, the soil and water loss area of China has reached 3.56 million km~2, accounting for 37% of the total land area of our country. The area of hydraulic erosion is 1.65 million km~2, and the area of wind erosion is 1.91 million km~2.. To our country's national production as well as the resident's life has caused the huge hidden danger and the threat. Due to the characteristics of soil and water loss, such as sudden, uncertain and catastrophic, accurate early warning of soil and water loss has become a worldwide scientific problem. Therefore, it is the most effective method to reduce the loss of soil and water loss. Remote sensing technology is a non-contact, new detection technology developed in 1960's. In general, we use remote sensing detection equipment to detect the reflected and scattering characteristics of the measured object. At the same time, combining with the characteristics of the related physical object, we study the new technology and means of the principle. Compared with the traditional manual detection technology, remote sensing technology has the following characteristics: large observation range, no time and space limitation, synthesis, macro and so on. It provides a powerful condition for the macroscopic study of various geographical phenomena and their interrelations, and the remote sensing data have the following characteristics: large amount of information and many means of obtaining, which makes researchers have the ability of all-weather and multi-directional observation of the earth; Fast information acquisition, short update period, dynamic detection and so on. Based on the above purpose, the main work and contents of this study are as follows: firstly, this paper introduces the research status of remote sensing technology applied to disaster information extraction, and the related concepts of parallel computing. The development course and the current mainstream parallel computing programming environment, the pattern has carried on the simple outline. On this basis, the application of parallel computing technology in remote sensing image processing is analyzed, and the process of building a hybrid programming model of MPI and OpenMP based on Berkerley NOW is introduced. The correlation coefficient method and the ratio method, which are commonly used in remote sensing image detection, are used as comparison information extraction methods. The parallel model based on master-slave mode is implemented on the cluster system in this study. Finally, the remote sensing image near Dashishan in Loudi area of Hunan Province is used as the experimental sampling area to carry on the contrast experiment, and the effect of parallel calculation is discussed.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP338.6;TP751

【参考文献】