本文選題：遙感影像數(shù)據(jù) + 海量�。� 參考：《河南大學》2015年碩士論文

【摘要】：針對遙感影像數(shù)據(jù),如生物物理參數(shù)產(chǎn)品、大型工程評價區(qū)生物環(huán)境變化監(jiān)測產(chǎn)品、國家級自然保護區(qū)植物優(yōu)勢群落監(jiān)測、國家級自然保護區(qū)生境破碎狀況預警產(chǎn)品、土壤遙感應用產(chǎn)品、區(qū)域生態(tài)環(huán)境災害遙感應用等進行生產(chǎn)、加工、處理時,第一步要對遙感影像數(shù)據(jù)進行分割目標,并通過目標識別來提取所需要的信息,第二步是將分割目標結果進行分析,并進行目標識別處理等,這個過程稱之為連通域標記。那么,為實現(xiàn)遙感影像數(shù)據(jù)進行連通域并行快速標記時怎么樣才能讓連通的區(qū)域之間的邊界呈現(xiàn)出光滑、自然的效果成為了一個疑問。因此,為了實現(xiàn)將海量遙感影像數(shù)據(jù)的連通區(qū)域邊界呈現(xiàn)出自然、光滑的效果,將原有的四向連通優(yōu)化為八向連通。但是,在處理一些數(shù)據(jù)量和計算量很大的海量遙感影像數(shù)據(jù)(即灰度圖像)時,目前現(xiàn)有算法將不能滿足該需求。為了解決該問題,研究一種針對海量遙感影像數(shù)據(jù)快速、有效地并行標記算法成為重點。為了解決以上問題,本文研究通過并行標記算法使其在海量遙感影像數(shù)據(jù)基礎上使用,并考慮怎么樣將該算法在“高分辨率對地觀測系統(tǒng)”的實際研發(fā)過程中得到應用。下面是對其主要內(nèi)容的研究和貢獻。(1)提出一種針對海量遙感影像數(shù)據(jù)八連通域并行標記算法。針對目前已有連通域標記算法在遙感影像數(shù)據(jù)八向連通域快速標記不適用問題在現(xiàn)有連通域標記算法的基礎上,將對傳統(tǒng)八向連通域標記算法從方向上進行優(yōu)化(即八向連通優(yōu)化為四向連通),在連通標記處理過程中的沖突問題時,采用列表沖突處理機制解決,同時利用分塊處理和合并處理機制,不僅實現(xiàn)海量遙感影像數(shù)據(jù)在連通域中對圖像的正確劃分,而且可以精確、有效地計算出連通域數(shù)目,并對海量遙感影像數(shù)據(jù)在八向連通域中有效地進行了并行標記的快速處理,從而得到一幅完整的處理后的海量遙感影像數(shù)據(jù)。通過實驗結果對比表明,本文提出的基于海量灰度圖像八連通域并行標記算法,不僅可以適用于目前的處理要求,而且在處理海量遙感影像數(shù)據(jù)中具有高效性。(2)國家重大專項課題之一——“高分辨率對地觀測系統(tǒng)”實際研發(fā)過程中,已經(jīng)將海量灰度圖像八連通域并行標記算法應用在其中。在實際的研發(fā)和設計的過程中,為了使“高分辨率對地觀測系統(tǒng)”能有效地解決對遙感影像數(shù)據(jù)進行分割目標識別和信息提取,成功地將海量灰度圖像八連通域并行標記算法應用在項目研發(fā)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.
【學位授予單位】：河南大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：TP751

【相似文獻】

相關期刊論文 前10條

1 左建政，樓志文，，張海濤;任意多連通域網(wǎng)格自動生成算法及其應用[J];計算結構力學及其應用;1995年01期

2 馬江林;趙忠明;孟瑜;彭玲;;海量遙感分類圖連通域標記方法[J];計算機工程;2008年01期

3 劉曉平;何士雙;;基于三角劃分的多連通域圖形匹配研究[J];工程圖學學報;2010年01期

4 樊大鈞;單和多連通區(qū)平面問題的近似解[J];兵工學報;1980年01期

5 黃新民;;圓環(huán)域到一類兩連通域的保角映射[J];廣西大學學報(自然科學版);1989年02期

6 郭麗,黃元元,楊靜宇;基于連通域的版面分割研究[J];南京理工大學學報(自然科學版);2003年01期

7 盧亞玲;;基于迭代法和連通域的儲糧圖像糧蟲分割算法[J];湖北工業(yè)大學學報;2007年06期

8 于泳波;李萬恒;張勁泉;聶建國;;基于圖像連通域的橋梁裂縫提取方法[J];公路交通科技;2011年07期

9 李儀芳;劉景琳;;基于連通域算法的區(qū)域測量[J];科學技術與工程;2008年09期

10 張凱兵;;基于連通域的字符孔洞提取算法及其實現(xiàn)[J];西華大學學報(自然科學版);2007年04期

相關會議論文 前2條

1 徐姍姍;李學明;;基于連通域合并的多運動目標跟蹤[A];第十四屆全國信號處理學術年會(CCSP-2009)論文集[C];2009年

2 黃磊;劉昌平;;信封圖像的快速傾斜矯正算法[A];第八屆全國漢字識別學術會議論文集[C];2002年

相關博士學位論文 前1條

1 周景超;視頻文本檢測算法研究[D];中國科學院研究生院（自動化研究所）;2008年

相關碩士學位論文 前10條

1 梅夢麗;基于背景連通域的印刷線路板缺陷定位及識別[D];南昌航空大學;2015年

2 黃自力;場景文字定位方法研究及實現(xiàn)[D];電子科技大學;2014年

3 陳慶慶;基于機器視覺的表具走字識別技術研究[D];北京工業(yè)大學;2015年

4 付龍;足球視頻中多目標跟蹤算法研究[D];河北工業(yè)大學;2015年

5 薛征;車牌識別系統(tǒng)關鍵技術的研究[D];東北石油大學;2015年

6 姜建飛;基于移動云計算的即時翻譯軟件的設計與實現(xiàn)[D];東南大學;2015年

7 范家銘;海量灰度圖像八連通域并行標記算法的研究與應用[D];河南大學;2015年

8 嵇新浩;基于連通域的文本定位方法研究[D];浙江工業(yè)大學;2007年

9 郭正東;基于連通域的扭曲文本圖像校正方法研究與應用[D];北方工業(yè)大學;2015年

10 王晶晶;海量灰度圖像8連通域標記算法的設計與應用[D];河南大學;2014年

本文編號：1888723