博碩士論文查詢聯邦系統首頁 聯絡信箱 關於我們 加入我們

本論文已被瀏覽 36 次, [ 造訪詳細資料與全文 ] 66 次,[ 回到前頁查詢結果 ] [ 重新搜尋 ]

應用MODIS資料進行土壤含水量乾旱與潮濕的分類

Using MODIS satellite data for Drought/Wetness classification of soil moisture contents

作者:吳俊龍
畢業學校:國立聯合大學
出版單位:國立聯合大學
核准日期:2010-09-14
類型:Electronic Thesis or Dissertation
權限:Copyright information available at source archive--National United University....

中文摘要

土壤含水量 (Soil Moisture Content, SMC)在對於水文 (hydrology)、農業 (agronomy) 和氣象 (meteorology) 方面佔有重要因素。因此,本研究利用MODIS衛星資料針對台灣北區地區的土壤含水量分為兩類型:乾旱 (Drought) 和潮濕 (Wetness)。我們提出兩層機率模型來估計土壤含水量的類型,而馬可夫隨機場估計高斯分佈是基於MODIS 資料的NDVI與LST代入回歸模式所構成。在這種模式下,底層是馬可夫隨機場 (MRF) 代表土壤水分含量的類型,而頂層是描述土壤含水量的分佈相依於底層的MRF可得高斯分佈,並估計的土壤含水量類型且逹到MODIS 影像分割的最大事後機率(MAP)。
實驗結果證明本研究方法是基於MRF與回歸 MODIS資料能夠成功地區分森林地區中的乾旱區域,而分類結果也符合中央氣象局所提供的實際降雨量。分別觀察不同的日期的回歸土壤含水量的趨勢是類似石門和寶山第二水庫觀測站實際測量土壤含水量的趨勢,另外我們的模型有兩個優點。首先,它可以成功地快速的分辨“乾旱”和“潮濕”區域。而這區別能力不見得可以用更為複雜的雙線性來源模型得到。其次,對於區域有雲層覆蓋因而缺乏MODIS資料去分類,利用我們的模型可成功的區分出為乾旱或潮濕。所以,若能即時掌控土壤含水量的狀況,即可用於進一步水文和乾旱管理與應用。

英文摘要

Soil Moisture Content (SMC) plays an important role in hydrology, agronomy, and meteorology. Our method based on MRF and regression of MODIS data can successfully distinguish drought and wetness. We propose two-layer statistical model to estimate the type of soil moisture content. This estimation is modeled as a Markov random field over which Gaussian variables are constructed from statistical analysis of regression of NDVI and LST MODIS data. Under this model, the bottom layer is the Markov random field (MRF) that represents the types of soil moisture content, the top layer describes the observations of soil moisture content whose distributions are Gaussians dependent of the underlying MRF realizations, and the estimation of SM types is achieved by the maximum a posteriori (MAP) classification of MODIS data.
Experimental results show that our method based on MRF and regression of MODIS data can successfully segment the wooded grassland region under studying. Our classification results are in line with the actual rainfall provided by the Central Weather Bureau. The trends in the SM observation along different dates are similar to those in the actual measurements at Shihmen and Bashan reservoir observatory. Our method has additionally two advantages. Firstly, it can successfully distinguish drought and wetness, while this distinguishing may not be achieved by the linear two-source model, which is much more complex. Secondly, the regions covered by the cloud and hence lacking MODIS data can be classified as drought or wetness in the classification. The type information resulted from our classification can be used for further applications in hydrology or drought management.


委員 - 何肯忠

委員 - 曾裕強

召集委員 - 楊世任


 

計畫贊助者: