摘要高光谱遥感技术在当今是非常热门的一个研究方向，高光谱图像的分类也是人们热衷的学术方向。但是由于高光谱的数据量越来越大，同时，分类算法也变得越来越复杂，传统的分类方法在如今已经不能满足人们对它的应用需求，所以迫切需要新的分类方法或者对传统方法进行有效的优化。基于GPU的并行优化算法相对传统方法来说，在分类效率上有质的飞跃。GPU具有强大的并行处理和浮点计算能力，适用于数据量大、计算密集型的运算，能够很好地解决以上问题。28896
本文在总结高光谱遥感技术和高光谱分类研究现状、分析比较传统高光谱分类方法的基础上，研究了基于空谱联合核的高光谱图像稀疏表示分类的GPU并行优化模型和算法，设计了对应的方法和流程，并在Visual Studio2010平台上开发与实现了基于GPU的高光谱遥感图像分类软件系统，包括了文件读取、图像分类、精度显示等功能，给出了系统核心模块以及算法的详细设计。最后通过实际高光谱数据进行了详细的比较试验，验证了本文方法和软件功能的有效性。
关键词 高光谱 GPU并行 分类
毕业论文设计说明书外文摘要
Title System Design and Implementation of Hyperspectral Remote Sensing Image Classification Based on GPU
Abstract
Hyperspectral remote sensing technology is a hot research direction in today, and the classification of hyperspectral image is also a hot academic direction.. But due to the increasingly large amount of hyperspectral data. At the same time, classification algorithm has become more and more complex, the traditional classification methods in today has been unable to meet the application requirements of the people on it, so there is an urgent need, a new method of classification or effective optimization to the traditional method. The parallel optimization algorithm based on GPU has a qualitative leap in classification efficiency.GPU parallel optimization algorithm can effectively reduce the data exchange between GPU and CPU, and can well solve the above problems.
Based on the summary of hyperspectral remote sensing technology and the status quo of the hyperspectral classification research, analysis, comparison of the traditional method of hyperspectral classification. Is studied on the basis of this, based on joint nuclear empty spectrum of hyperspectral image sparse representation classification of GPU parallel optimization model and algorithm, designed the corresponding method and process, and in Visual Studio2010 platform development and implementation of the hyperspectral remote sensing image classification based on GPU software system, including the file read, image classification, display the function such as precision, gives the system and the core modules and the detailed design of the algorithm. At last, through actual hyperspectral data has carried on the detailed comparison test, this method was verified and software function effectiveness.
Keywords hyperspectral; GPU parallel; classification
目 次
1 绪论 1
1.1 课题背景 1
1.2 研究意义 3
1.3 研究现状 3
1.4 论文主要内容 5
2 相关概念与技术 6
2.1高光谱图像分类 6
2.2 Visual stuidio 2010和MFC 6
2.3基于CUDA的GPU并行计算 7
3 基于空谱联合核稀疏表示分类的GPU并行优化 9
3.1 基于空谱联合核稀疏表示分类算法原理 9
3.2 基于GPU的并行优化设计 12
4 系统设计与实现 15
4.1系统需求分析 15
4.2系统设计 15 基于GPU的高光谱图像遥感分类系统设计与实现:http://www.youerw.com/jisuanji/lunwen_23918.html