【作者】 李振华；

【导师】 文灏；

【作者基本信息】 华中科技大学 ， 电路与系统， 2011， 硕士

【摘要】 高光谱成像技术是在多光谱的技术之上发展而来,相对于多光谱图像,高光谱图像具有更高的光谱分辨率以及更多的波段数,因此也带来了更丰富的信息。高光谱图像的处理与分析技术在遥感技术的发展过程中起着重要的作用。目前,国内外在高光谱图像分析技术上已取得了丰富的成果,每天都有大量的新方法出现,但是,真正需要将这些技术推广到实际的应用中,就必须要有专业的图像处理软件。本文主要对高光谱图像处理平台的设计进行了研究,详细阐述了平台的设计过程,以及分析了在平台的开发过程中遇到的关键技术问题,并给出了相应的解决方法。同时,在图像的非监督分类算法上进行了软件化实现。本论文的研究重点包括平台的设计与研究,分析了高光谱图像处理平台的主要功能,同时针对这些主要功能进行了模块的划分,详细的介绍了各个模块的主要功能和设计过程,并且设计了模块之间相互联系的接口。对平台设计过程中遇到的关键技术难题进行了分析,针对相应的问题提出了解决方案。并以图像的非监督分类为实例,利用平台提供的接口进行了算法的实现,并针对K均值和IsoData算法初始类别中心的选择的方法进行了改进,而且提供了在IsoData算法中分裂和合并的阈值选取的方法。本论文的主要工作是提供了进行图像处理算法开发的接口,为高光谱图像处理平台的扩展提供了基础。论文的最后,对平台实现的功能进行了相应的测试,对比了与ENVI软件的操作结果,并总结了目前平台的特点与不足之处。更多还原

【Abstract】 Hyperspectral imaging techiques was developed from multispectral techiques.As opposed to the multilspectral image,the hyperspectral image has a higher spectral resolution and more bands,and therefore brings a richer information.The processing and analysis techiques of Hyperspectral image played a role in the development of the remote sensing thchnology.A large number of achievement was gained both at home and aboard nowadays,and lots of new methods was arose everyday.But,it is necessary to develop a professional image processing software to promote these techniques to practical applications.This paper has studied the design of hyperspectral image processing platform,and makes expatiation of the design procedure of platform,and some key techniques which was encountered during the design procedure was analysised,and The resolution was presented according the key techniques.Meanwhile,unsupervised classification algorithm for image processing has been implement by software.The research of this papaer focused on the design and research of the paltorm,and analyzed the main functions of the platform,at the same time,the modules was divided according to the main functions, described in detail the main function of each module and the design process,designed the contact interface with each other of the module. The key technical problems encountered on the design process were analyzed, and offered corresponding resolution.And take the example of non-supervised classification of image to realize the algorithm using the interface of the platform,imporved the method of the choice of initial centers for algorithm,and also provided the mothod that the threshold was choosed of the split and merge process for the IsoData algorithm.The main work of this paper provided the interface for the development of image processing algorithm,and laid the foundations of extensions for the hyperspectral image processing platform.The last part were the test for the realization of the platform, the results compared with the ENVI software,and the conclusions of the chief characteristics and defects.更多还原