面向森林变化监测的海量遥感数据并行处理技术研究

随着遥感技术的发展和遥感影像林业应用的深入，应用中高空间分辨率遥感影像开展森林类型精细识别和森林资源动态变化监测成为目前研究的热点之一。随着对地观测技术的快速发展，尤其是高分辨率对地观测时代的来临，遥感图像呈数据的爆炸式增长，如何更好的利用海量的遥感影像数据，近实时的解决森林变化监测中各类应用，已成为近年来研究的热点。本项研究面向森林资源变化监测中的海量遥感图像准确、高效实时处理要求，从“二点一线”展开研究，两点分别是指针基于Spark的并行编程模型，该模型通过提高并行编程的抽象层次和支持的编程范式来实现MapReducede的实例化，隐式实现并行细节以及建立的分布式内存全局数据存储模型和面向SuperMap应用的分布式遥感图像存储框架，而一线是指面向森林资源动态变化的遥感数据密集型并行计算的量化模型。

With the development of remote sensing technology and applicating the remote sensing image in forestry application in depth, it has becoming one of the hot spot of current research to carry out forest type fine-grained identification and forest resource dynamic monitoring with applicating medium and high spatial resolution remote-sensing image. With the rapid development of earth observation technology, especially the coming of the era of high resolution earth observation, the remote sensing images data presents explosive growth, how to use of remote sensing image data better, as well as various kinds of near-real-time applications in solving the forest change monitoring, have become a research hotspot in recent years. The study in order to satisfies(满足) requirements of accurate and efficient real-time processing the mass remote sensing images in forest resources changes monitoring, we begin the study from the "two points one line", while "two points" are parallel programming model with pointers based on Spark, the model realizes the instantiation of MapReduce by improving the abstraction level of parallel programming and programming paradigm which is supported, implicitly realizes parallel details and establish the global data storage model of distributed memory as well as SuperMap application oriented distributed remote sensing image storage framework, and "A line" refers to the dynamic changes of forest resources oriented the quantitative model of remote sensing data-intensive parallel computing