English  |  正體中文  |  简体中文  |  Items with full text/Total items : 28611/40649
Visitors : 635228      Online Users : 73
RC Version 4.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Adv. Search
LoginUploadHelpAboutAdminister

Please use this identifier to cite or link to this item: http://ntour.ntou.edu.tw:8080/ir/handle/987654321/18992

Title: 高頻譜影像目標偵測、分類和壓縮之平行演算法則研究
Parallel Algorithm Approaches to Target Detection, Classification and Compression for Hyperspectral Images
Authors: 張麗娜;張陽郎
Contributors: NTOU:Department of Communications Navigation and Control Engineering
國立臺灣海洋大學:通訊與導航工程學系
Keywords: 高頻譜影像;目標偵測;影像分類;影像壓縮;訊號子空間投影法;部分濾波;平行計算系統
hyperspectral images;target detection;image classification;image compression;signal subspace projection;partial filters;parallel computing system
Date: 2008-08
Issue Date: 2011-08-17T07:05:54Z
Publisher: 行政院國家科學委員會
Abstract: 摘要:高頻譜影像具有更多的光譜波段數目和較高的空間及光譜解析度,因此比多頻譜影像提供了更豐富及細緻的地表資訊,對於目標偵測與影像分類也有較好的辨識能力。但是高頻帶數和高解析度也意味著高資料量,如何在龐大的資料量裡,發展一高效的目標物辨識和影像分類技術,就變得相當重要。本計畫第一年的研究乃針對高頻譜影像,提出以訊號子空間投影法為基礎之高效目標偵測研究。為了提昇高頻譜影像目標偵測的效能,我們依據高頻譜影像頻帶具有高相關的特性,提出部份濾波法來減少目標偵測的計算複雜度。首先提出訊號子空間投影法(Signal Subspace Projection, SSP)進行影像目標偵測研究,此法以高頻譜影像所對應的訊號子空間為基礎,透過子空間轉換萃取重要的特徵資訊,並以對應特徵為基礎設計最佳的濾波器,進行影像目標偵測。為了進一步降低SSP法的計算複雜度,並充分利用高頻譜影像頻帶間的高相關性,我們提出一結合部分濾波和訊號子空間的投影法(SSP) 以減少頻帶數量達到降低原來濾波器設計的計算量。此技術先以聚類頻帶分群法將相關性高的頻帶快速的聚集為一群組。接著,將高頻譜影像分群處理,針對每一群組,以SSP法設計最佳的部份濾波器,再結合各群組的部份濾波器進行目標偵測。由於部份濾波法具可平行處理的結構,我們以個人電腦建構叢集式平行計算系統來實現部份濾波法,以進一步提昇高頻譜影像處理的效能。模擬結果顯示,本研究所提部份濾波法,除了可減少原來濾波器法於高頻譜影像處理的計算量,也提高了目標偵測和影像分類的效能。此外,藉由建置的電腦叢集進行平行處理除了可快速驗證部份濾波法的效能,並明顯的提昇高頻譜影像於目標偵測的處理速度。
abstract:The hyperspectral imagery offers a better identification in target detecting and image classifying than multispectral imagery. However, higher spectral resolution increases the computation complexity in the target detection processing. In the first year of the study, to improve the efficiency in target detection and image classification, we propose a combined partial filter and signal subspace projection approach which exploits high degree correlations in spectral domain.First, a novel signal subspace projection (SSP) approach is proposed to detect and extract target signatures in unknown background. The SSP approach designs the weights of adaptive filter by using the signal subspace components corresponding to the image correlation matrix. The proposed SSP filter can be designed to extract or detect target signatures without priori knowledge of signatures or background required. Furthermore, to reduce the computation complexity of SSP, we proposed a combined SSP and partial filter (CSSPPF). We examine the correlations between hyperspectral image bands and partition the original image bands into several groups according to their correlations. Bands with highly correlated features are grouped together. We then design a partial filter to extract the corresponding signatures for each image group by SSP approach. Then, the SSP partial filters are combined to be a target detector, called a combined SSP and partial filter (CSSPPF). Since the structure of the partial filter is very suitable for the parallel processing, we then build a PC clustered parallel computing system to realize the CSSPPF approach. The PC cluster is constructed by four same personal computers and one server computer through the router and the network cable. And we use the Parallel Computing Toolbox provided by Matlab and the Matlab Distributed Computing Engine to execute the parallel process of the partial filter through the Distributed Computing. As shown in the experimental results, the proposed CSSPPF can eliminate the interference efficiently and improve the accuracy of the target detection with less computation complexity than conventional target detection schemes. In addition, the PC clustered parallel computing system fast validates the efficiency of the partial filter and speeds up the processing rate of target detection in hyperspectral images.
Relation: NSC97-2221-E019-011
URI: http://ntour.ntou.edu.tw/handle/987654321/18992
Appears in Collections:[通訊與導航工程學系] 研究計畫

Files in This Item:

There are no files associated with this item.



All items in NTOUR are protected by copyright, with all rights reserved.

 


著作權政策宣告: 本網站之內容為國立臺灣海洋大學所收錄之機構典藏,無償提供學術研究與公眾教育等公益性使用,請合理使用本網站之內容,以尊重著作權人之權益。
網站維護: 海大圖資處 圖書系統組
DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback