| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊工程學系 | zh_TW |
| DC.creator | 王詠令 | zh_TW |
| DC.creator | Yung-Ling Wang | en_US |
| dc.date.accessioned | 2014-6-19T07:39:07Z | |
| dc.date.available | 2014-6-19T07:39:07Z | |
| dc.date.issued | 2014 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=945402019 | |
| dc.contributor.department | 資訊工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 光學遙測利用物質的反射特性進行辨識，因為不同物質具有獨特的吸收帶而形成一種獨特的光譜特徵，運用此獨特性可以根據光譜辨別不同的物質來進行分類。傳統光譜辨別的方式，是直接測量光譜之間的距離或角度作為相似度分析，然而實際的光譜通常包含雜訊的干擾，傳統的測量方法沒有足夠的容錯能力而造成誤差。本研究提出一種新的方法來測量光譜辨別物質之間的相似性，我們採用經驗模式分解來將光譜分解成幾個本質分量，並使用這些分量以提高光譜辨識的性能。從分解的分量中發現，訊號與雜訊被區分在不同的分量，而吸收區資訊分散於前面數個分量中，這些分量具有更好的能力來辨別物質。為了方便評估，我們提出幾種常用的測量的方法來進行性能比較分析，如歐氏距離、光譜角度和馬氏距離。本實驗的樣本光譜是由美國地質調查局（USGS）的光譜庫提供，實驗結果證明經驗模式分解後的光譜相似性測量，能更有效地萃取光譜特徵，提升分類準確性。 | zh_TW |
| dc.description.abstract | Optical remote sensing can distinguish different materials because each material has its own unique absorption characteristics to form a unique spectrum. This information can be adopted to discriminate different materials in optical remote sensing images. Traditional approach for spectra similarity measurement is calculating the Euclidean distance or spectral angle between two spectra directly. However, in reality the spectra usually contain noise or interference which cannot be tolerated by traditional measurements. In this study, we propose a new approach to measure the similarity between the spectra to discriminate materials. It adopts Empirical Mode Decomposition (EMD) to decompose the spectrum into several components, called Intrinsic Mode Functions (IMFs). The absorption features are highlighted and the noise is reduced in the first few IMFs, so the ability of material discrimination is improved. For evaluation purpose, we compare the proposed method with several commonly used measurements, including Euclidean distance, Spectral Angle and Mahalanobis distance. The sample spectra used for experiment are provided by the spectral library of U. S. Geological Survey (USGS). The experiments results have demonstrated that EMD can extract the spectral features more effectively than common spectral similarity measurements and improve the classification performance. | en_US |
| DC.subject | 高光譜 | zh_TW |
| DC.subject | 經驗模式分解 | zh_TW |
| DC.subject | 歐氏距離 | zh_TW |
| DC.subject | 光譜角度 | zh_TW |
| DC.subject | 馬氏距離 | zh_TW |
| DC.subject | Hyperspectrum | en_US |
| DC.subject | Empirical Mode Decomposition (EMD) | en_US |
| DC.subject | Euclidean distance | en_US |
| DC.subject | Spectral Angle | en_US |
| DC.subject | Mahalanobis distance | en_US |
| DC.title | 經驗模式分解應用於高光譜資料分析 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Empirical Mode Decomposition for Hyperspectral Data Analysis | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |