海量灰度图像八连通域并行标记算法的研究与应用

发布时间：2018-05-14 16:57

本文选题：遥感影像数据 + 海量　；参考：《河南大学》2015年硕士论文

【摘要】：针对遥感影像数据,如生物物理参数产品、大型工程评价区生物环境变化监测产品、国家级自然保护区植物优势群落监测、国家级自然保护区生境破碎状况预警产品、土壤遥感应用产品、区域生态环境灾害遥感应用等进行生产、加工、处理时,第一步要对遥感影像数据进行分割目标,并通过目标识别来提取所需要的信息,第二步是将分割目标结果进行分析,并进行目标识别处理等,这个过程称之为连通域标记。那么,为实现遥感影像数据进行连通域并行快速标记时怎么样才能让连通的区域之间的边界呈现出光滑、自然的效果成为了一个疑问。因此,为了实现将海量遥感影像数据的连通区域边界呈现出自然、光滑的效果,将原有的四向连通优化为八向连通。但是,在处理一些数据量和计算量很大的海量遥感影像数据(即灰度图像)时,目前现有算法将不能满足该需求。为了解决该问题,研究一种针对海量遥感影像数据快速、有效地并行标记算法成为重点。为了解决以上问题,本文研究通过并行标记算法使其在海量遥感影像数据基础上使用,并考虑怎么样将该算法在“高分辨率对地观测系统”的实际研发过程中得到应用。下面是对其主要内容的研究和贡献。(1)提出一种针对海量遥感影像数据八连通域并行标记算法。针对目前已有连通域标记算法在遥感影像数据八向连通域快速标记不适用问题在现有连通域标记算法的基础上,将对传统八向连通域标记算法从方向上进行优化(即八向连通优化为四向连通),在连通标记处理过程中的冲突问题时,采用列表冲突处理机制解决,同时利用分块处理和合并处理机制,不仅实现海量遥感影像数据在连通域中对图像的正确划分,而且可以精确、有效地计算出连通域数目,并对海量遥感影像数据在八向连通域中有效地进行了并行标记的快速处理,从而得到一幅完整的处理后的海量遥感影像数据。通过实验结果对比表明,本文提出的基于海量灰度图像八连通域并行标记算法,不仅可以适用于目前的处理要求,而且在处理海量遥感影像数据中具有高效性。(2)国家重大专项课题之一——“高分辨率对地观测系统”实际研发过程中,已经将海量灰度图像八连通域并行标记算法应用在其中。在实际的研发和设计的过程中,为了使“高分辨率对地观测系统”能有效地解决对遥感影像数据进行分割目标识别和信息提取,成功地将海量灰度图像八连通域并行标记算法应用在项目研发中。
[Abstract]:For remote sensing image data, such as biophysical parameter products, biologic environment change monitoring products in large engineering evaluation areas, plant dominant communities monitoring in national nature reserves, and early warning products for habitat fragmentation in national nature reserves. In the production, processing and processing of soil remote sensing application products and remote sensing applications for regional ecological environmental disasters, the first step is to segment the remote sensing image data and extract the needed information through target recognition. The second step is to analyze the result of segmentation and process target recognition, which is called connected domain marking. So, how to make the boundary between connected regions smooth and smooth in order to realize the connected domain parallel and fast marking of remote sensing image data, the natural effect has become a question. Therefore, in order to realize the natural and smooth boundary of the connected region of the massive remote sensing image data, the original four-direction connectivity is optimized to eight-direction connectivity. However, when dealing with some massive remote sensing image data (i.e. gray-scale image) with large amount of data and computation, the existing algorithms will not meet this requirement. In order to solve this problem, a fast and efficient parallel marking algorithm for massive remote sensing image data has become the focus. In order to solve the above problems, this paper studies how to use the parallel marking algorithm on the basis of massive remote sensing image data, and consider how to apply the algorithm in the development of "High Resolution Earth observation system". The following is the research and contribution of its main content. 1) A parallel tagging algorithm for eight connected domains of massive remote sensing image data is proposed. In view of the problem of fast tagging of connected domain in remote sensing image data, the existing connected domain labeling algorithms are based on the existing connected domain labeling algorithms. In this paper, the traditional eight-direction connected domain labeling algorithm is optimized in the direction (that is, the octagonal connectivity is optimized as the four-direction connectivity). When dealing with the conflict problem in the process of the connected label processing, the list conflict processing mechanism is used to solve the problem. At the same time, using the mechanism of block processing and merging processing, not only can the massive remote sensing image data be divided correctly in connected domain, but also the number of connected domain can be calculated accurately and effectively. The large amount of remote sensing image data is processed efficiently in the octadirectional connected domain, and a complete processed mass remote sensing image data is obtained. The experimental results show that the proposed parallel marking algorithm based on eight connected domains of massive gray images can not only be applied to the current processing requirements. Moreover, in the process of practical research and development of "High Resolution Earth observation system", one of the most important national special projects, "High Resolution Earth observation system", which has high efficiency in processing massive remote sensing image data, has been applied to the eight connected domain parallel marking algorithm of massive gray scale image. In the process of actual research and design, in order to make "high-resolution Earth observation system" can effectively solve the segmentation of remote sensing image data target recognition and information extraction, Eight connected domain parallel marking algorithms of massive gray images are successfully applied in project development.
【学位授予单位】：河南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP751

【相似文献】