CUHK electronic theses & dissertations collection

生長於潮間帶的紅樹林是熱帶和亞熱帶地區最具生產力的生態系統之一。香港擁有十個紅樹品種，其覆蓋面積約共三百五十公頃。位於香港西北面的米埔是現時香港最大的紅樹林區。這片紅樹林及其鄰近濕地於一九九五年被列為拉姆薩爾重要的濕地。隨著經濟的迅速發展、污染及一些不可持續的開發，全球紅樹林的面積不斷地萎縮。而香港的紅樹也正面對城市發展及基建的直接威脅。因此，了解及監測紅樹林的生長狀況、覆蓋面積的轉變是紅樹林保育的基礎。遙感是具有成本效益和能提供及時數據的技術，在紅樹林的生態保育及監測上發揮著重要功能。是次研究選擇位於米埔的紅樹林區。通過結合高光譜和雷達數據以及實地磡測，以達到三個目的。第一，利用模式辨認分析找出可提高品種辨識度的光譜帶及雷達數據。第二，把挑選出來的光譜帶及雷達數據組合，利用不同的分類法包括最大概似法、决策樹 C5.0演算法、類神經網路及支持向量機進行紅樹林的品種分類，並籍此測試各分類法的精度。第三，利用植被指數及雷達數據中取得的參數為獨立變量，而在野外點測的葉面積指數 (LAI) 為因變量，通過迴歸分析以估算整片紅樹林的葉面積指數，籍此了解紅樹林現時的生物物理狀況。根據特徵選擇的結果，位於高光譜數據中的綠波段 (570nm, 580nm, 591nm及601nm)、紅波段 (702nm)、紅邊位 (713nm)、近紅外波段 (764nm及774nm)、 短波紅外波段 (1276nm, 1316nm及1629nm) 以及在不同季節取得的過濾後向散射數據是最能辨識品種差異。據品種分類的結果顯示，單用多時後向散射特徵數據存在很大誤差。而在大多的情況下，單用光譜數據比起混合光譜及後向散射數據的分類表現為佳。但對於某些品種來說，後向散射數據能給予比較準確的預測。另外，在同數據組合下，分類法在訓練精度上沒有多大的分別。除了類神經網路分類法以外，其他分類法的測試精度總比其訓練精度低。這說明類神經網路模型比起其他分類法的模型要為穩定，而决策樹模型則被過度訓練。根據生產者及使用者精度分析，因為缺乏足夠的訓練樣本，桐花樹及海桑屬的精度較其他品種為低。據不同植被指數的簡單線性迴歸模型顯示，利用三角植被指數 (TVI)及修正葉綠素吸納比例指數一 (MCARI 1) 對於葉面積指數的估算是最準確。相反地，葉面積指數與從雷達數據中取得的參數關係則比較弱。這表示單用雷達參數不能對葉面積指數進行準確的估算。在結合植被指數及雷達參數的多元逐步迴歸分析下，三角植被指數及在灰度共生矩陣下得出的角二階矩參數能減低葉面積指數估算的誤差。總結以上兩項分析，光譜及雷達數據在紅樹林的品種分類及葉面積指數估算上有互補的作用。Mangrove is one of the most productive ecosystems flourished in the intertidal zone of tropical and subtropical regions. Hong Kong has ten true mangrove species covering an approximate area of 350 hectares. Mai Po locating in the northwestern part of Hong Kong nourishes the largest mangrove stand and it was listed as a Wetland of Importance under the Ramsar Convention in 1995. Over the years, areas of mangrove have been shrinking globally due to development, pollution, and other unsustainable exploitation and Hong Kong was no exception. In Hong Kong, mangroves are usually sacrificed for urban development and infrastructure construction. Therefore, it is crucial to monitor their growth conditions, change of extent and possible unsustainable practices threatening their existence. Remote sensing being a cost-effective and timely tool for vegetation conservation is most suitable for such purpose.Taking Mai Po as study area, this study acquired satellite-borne hyperspectral and radar data supplemented with in situ field survey to achieve three purposes. First, features from the remotely-sensed data that are significant to species discrimination were identified through pattern recognition. Second, selected features grouped into different subsets were used to delineate the boundary of mangrove species through supervised classification. In the meantime, classifiers including maximum likelihood (ML), decision tree C5.0 (DT), artificial neural network (ANN) and support vector machines (SVM) were tested for their accuracy performance. The third purpose is to understand the current biophysical condition of mangrove through leaf area index (LAI) modeling by regressing field-measured LAI against vegetation indices, backscatter and textural measures.Results from feature selection revealed that hyperspectral narrowbands locating in green at 570nm, 580nm, 591nm, 601nm; red at 702nm; red-edge at 713nm; near infrared at 764nm and 774nm and shortwave infrared at 1276nm, 1316nm and 1629nm as well as the multi-temporal filtered backscatter captured in different seasons have high sensitivity to species difference.Species-based classification using multi-temporal backscatter features alone do not provide a satisfactory accuracy. Comparatively, results from pure spectral bands have better overall accuracy than that from combining spectral and radar features. However, radar backscatter does improve accuracy of some species. Besides, all classifiers had similar variations of training accuracy under the same feature subset. However, the testing accuracy is much lower with the exception of ANN. Performance of ANN was more stable and robust than other classifiers while serious overtraining occurs for the DT classifier. Moreover, most species were mapped accurately as revealed by the producer’s and user’s accuracy with the exception of A. corniculatum and Sonneratia spp. due to deficiency of training samples.Simple linear regression model with VIs revealed that triangular vegetation index (TVI) and modified chlorophyll absorption ratio index 1 (MCARI1) had the best relationship with LAI. However, weak relationship was found between field- measured LAI and radar parameters suggesting that radar parameters cannot be used as single predictor for LAI. Results from stepwise multiple regression suggested that TVI combined with GLCM-derived angular second moment (ASM) can reduce the estimation error of LAI. To conclude, the study has demonstrated spectral and radar data are complementarity for accurate species discrimination and LAI mapping.Detailed summary in vernacular field only.Detailed summary in vernacular field only.Detailed summary in vernacular field only.Detailed summary in vernacular field only.Detailed summary in vernacular field only.Wong, Kwan Kit.Thesis (Ph.D.)--Chinese University of Hong Kong, 2012.Includes bibliographical references (leaves 434-472).Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.Abstract also in Chinese.ACKNOWLEDGEMENTS --- p.IIABSTRACT --- p.IV論文摘要 --- p.VITABLE OF CONTENTS --- p.VIIILIST OF ABBREVIATIONS --- p.XIIILIST OF TABLES --- p.XVLIST OF FIGURES --- p.XVIIIChapter CHAPTER 1 --- INTRODUCTION --- p.1Chapter 1.1 --- BACKGROUND TO THE STUDY --- p.1Chapter 1.1.1 --- Mangrove Mapping and Monitoring --- p.1Chapter 1.1.2 --- Mangrove Mapping and Monitoring --- p.3Chapter 1.1.3 --- Role of Remote Sensing in Mangrove Study --- p.4Chapter 1.2 --- OBJECTIVES OF THE STUDY --- p.6Chapter 1.3 --- SIGNIFICANCE OF THE STUDY --- p.7Chapter 1.4 --- ORGANIZATION OF THE THESIS --- p.8Chapter CHAPTER 2 --- LITERATURE REVIEW --- p.10Chapter 2.1 --- INTRODUCTION --- p.10Chapter 2.2 --- FACTORS AFFECTING VEGETATION REFLECTANCE --- p.11Chapter 2.2.1 --- Foliar structure and principal constituents --- p.12Chapter 2.2.2 --- Foliar optical properties --- p.14Chapter 2.2.2.1 --- The visible region (400-700nm) --- p.14Chapter 2.2.2.2 --- The red edge (690-740nm) --- p.15Chapter 2.2.2.3 --- The near-infrared region (700-1300nm) --- p.16Chapter 2.2.2.4 --- The short-wave infrared region (1300-2500nm) --- p.17Chapter 2.2.3 --- Canopy architecture --- p.18Chapter 2.2.4 --- Background reflectance --- p.19Chapter 2.2.5 --- Atmospheric perturbation --- p.20Chapter 2.2.6 --- Sun-sensor relationship --- p.22Chapter 2.3 --- HYPERSPECTRAL IMAGING AND VEGETATION CLASSIFICATION --- p.23Chapter 2.4 --- RADAR IMAGING AND VEGETATION CLASSIFICATION --- p.31Chapter 2.5 --- PATTERN RECOGNITION FOR VEGETATION CLASSIFICATION --- p.39Chapter 2.5.1 --- The Hughes Phenomenon and Dimensionality Reduction --- p.39Chapter 2.5.2 --- Statistical Pattern Recognition and Feature Selection --- p.44Chapter 2.5.2.1 --- Search Method --- p.47Chapter 2.5.2.1.1 --- Exhaustive search --- p.48Chapter 2.5.2.1.2 --- Branch and bound --- p.49Chapter 2.5.2.1.3 --- Sequential forward/ backward selection --- p.55Chapter 2.5.2.1.4 --- Sequential Floating search --- p.57Chapter 2.5.2.1.5 --- Oscillating Search --- p.61Chapter 2.5.2.1.6 --- Genetic algorithm --- p.64Chapter 2.5.2.2 --- Evaluation criteria --- p.66Chapter 2.5.2.2.1 --- Distance measure --- p.67Chapter 2.5.2.2.2 --- Information measure --- p.68Chapter 2.5.2.2.3 --- Classification error --- p.71Chapter 2.5.2.3 --- Feature Selection Stability --- p.72Chapter 2.5.3 --- Feature extraction --- p.75Chapter 2.6 --- BIOPHYSICAL PARAMETERS MEASUREMENT AND ESTIMATION --- p.77Chapter 2.6.1 --- Leaf Area Index (LAI) --- p.78Chapter 2.6.2 --- Fraction of Absorbed Photosynthetically Active Radiation (fAPAR) --- p.79Chapter 2.6.3 --- In-situ Leaf Area Index Measurement --- p.81Chapter 2.6.3.1 --- Direct and Indirect Methods --- p.81Chapter 2.6.3.2 --- LAI Estimation through Gap Fraction Inversion --- p.85Chapter 2.6.3.3 --- Gap Fraction Ground Measurement --- p.89Chapter 2.6.3.3.1 --- LAI-2000 Plant Canopy Analyzer --- p.89Chapter 2.6.3.3.2 --- Hemispherical Photography --- p.92Chapter 2.6.3.4 --- Correction of Indirect LAI Measurement --- p.99Chapter 2.6.3.4.1 --- Clumping --- p.100Chapter 2.6.3.4.2 --- Mixture of Green and Non-green Elements --- p.101Chapter 2.6.4 --- Empirical Relationship with Spectral Vegetation Indices --- p.102Chapter 2.6.4.1 --- Traditional Vegetation Indices --- p.103Chapter 2.6.4.2 --- Leaf Area Index Estimation from Hyperspectral and Radar Images --- p.106Chapter 2.6.5 --- Physically-based Canopy Reflectance Model Inversion --- p.111Chapter 2.6.5.1 --- Canopy Reflectance Model --- p.111Chapter 2.6.5.2 --- Model Inversion and Biophysical Parameters Extraction --- p.115Chapter 2.7 --- SUMMARY --- p.118Chapter CHAPTER 3 --- METHODOLOGY --- p.120Chapter 3.1 --- INTRODUCTION --- p.120Chapter 3.2 --- STUDY AREA DESCRIPTION --- p.120Chapter 3.3 --- METHODOLOGICAL FLOW --- p.124Chapter 3.4 --- REMOTE SENSING DATA ACQUISITION AND PROCESSING --- p.127Chapter 3.4.1 --- Hyperion - EO-1 --- p.127Chapter 3.4.1.1 --- Radiometric correction --- p.127Chapter 3.4.1.1.1 --- Vertical strips removal --- p.128Chapter 3.4.1.1.2 --- Atmospheric correction --- p.129Chapter 3.4.1.1.3 --- Wavelength recalibration --- p.135Chapter 3.4.1.1.4 --- SNR enhancement through MNF --- p.137Chapter 3.4.1.2 --- Geometric correction --- p.139Chapter 3.4.1.3 --- Atmospheric correction algorithms comparison --- p.140Chapter 3.4.2 --- ASAR - ENVISAT --- p.141Chapter 3.4.2.1 --- Data Acquisition --- p.141Chapter 3.4.2.2 --- Data Processing --- p.143Chapter 3.4.2.2.1 --- Radiometric and Geometric Correction --- p.145Chapter 3.4.2.2.2 --- Speckle Filtering --- p.146Chapter 3.5 --- FIELD MEASUREMENTS AND DATA PROCESSING --- p.149Chapter 3.5.1 --- Species Distribution --- p.149Chapter 3.5.2 --- Leaf Spectra Measurement --- p.151Chapter 3.5.2.1 --- Leaf Collection and Handling --- p.152Chapter 3.5.2.2 --- ASD FieldSpec 3 Setup --- p.154Chapter 3.5.2.3 --- Laboratory setup --- p.156Chapter 3.5.2.4 --- Spectra Measurement --- p.158Chapter 3.5.2.5 --- Spectral similarity and variability --- p.159Chapter 3.5.3 --- In situ Leaf Area Index Measurement --- p.161Chapter 3.5.3.1 --- The optical instrument --- p.161Chapter 3.5.3.2 --- The LAI survey campaign p163Chapter 3.5.3.3 --- Data processing and canopy analysis --- p.166Chapter 3.5.3.4 --- Canopy parameter computation gap fraction, LAI, clumping index, mean inclination angle --- p.170Chapter 3.5.3.5 --- Field LAI and Their Correlation with Reflectance and Backscattering Coefficient Data Exploration --- p.175Chapter 3.6 --- FEATURE SELECTION --- p.175Chapter 3.6.1 --- Data Preprocessing and Preparation --- p.178Chapter 3.6.2 --- Data Format and Split --- p.183Chapter 3.6.3 --- Wrapper-based Approach --- p.185Chapter 3.6.4 --- Search Algorithm --- p.187Chapter 3.6.5 --- Stability Evaluation --- p.187Chapter 3.6.6 --- Feature Frequency analysis --- p.188Chapter 3.7 --- MANGROVE SPECIES CLASSIFICATION --- p.189Chapter 3.7.1 --- Species Separability --- p.193Chapter 3.7.2 --- Gaussian Maximum Likelihood Classifier --- p.193Chapter 3.7.3 --- Decision Tree Classifier --- p.194Chapter 3.7.4 --- Artificial Neural Network Classifier --- p.197Chapter 3.7.5 --- Support Vector Machines Classifier --- p.199Chapter 3.7.6 --- Accuracy Assessment --- p.204Chapter 3.8 --- LEAF AREA INDEX MODELING --- p.206Chapter 3.8.1 --- Preliminary Exploration of Relationship between Hyperspectral bands and LAI --- p.206Chapter 3.8.2 --- Vegetation Index Derived from Hyperspectral Data. --- p.206Chapter 3.8.3 --- Radar Backscatter and Derived Textural Parameters --- p.208Chapter 3.8.4 --- Regression Analysis --- p.211Chapter 3.8.5 --- Error Estimation --- p.217Chapter 3.9 --- SUMMARY --- p.218Chapter CHAPTER 4 --- RESULTS AND DISCUSSION (I) FEATURE SELECTION AND MANGROVE SPECIES CLASSIFICATION --- p.221Chapter 4.1 --- INTRODUCTION --- p.221Chapter 4.2 --- DATA PROCESSING AND EXPLORATION --- p.221Chapter 4.2.1 --- Atmospheric correction algorithms comparison --- p.222Chapter 4.2.2 --- Radar Data Speckle Reduction --- p.227Chapter 4.2.3 --- Statistical Discrimination of Mangrove Spectral Class --- p.230Chapter 4.3 --- FEATURE SELECTION --- p.249Chapter 4.3.1 --- Sequential Forward Selection (SFS) --- p.250Chapter 4.3.2 --- Sequential Floating Forward Selection (SFFS). --- p.256Chapter 4.3.3 --- Oscillating Search (OS) --- p.262Chapter 4.3.4 --- Search Algorithms comparison --- p.268Chapter 4.3.5 --- Final Subset Selection --- p.270Chapter 4.3.6 --- Correlation Analysis --- p.280Chapter 4.4 --- IMAGE CLASSIFICATION --- p.283Chapter 4.4.1 --- Mangrove Spectral Class Separability --- p.284Chapter 4.4.2 --- Gaussian Maximum Likelihood (ML) --- p.288Chapter 4.4.3 --- Decision Tree (DT) --- p.297Chapter 4.4.4 --- Artificial Neural Network (ANN) --- p.304Chapter 4.4.5 --- Support Vector Machines (SVM) --- p.312Chapter 4.4.6 --- Algorithm Comparison --- p.321Chapter 4.5 --- DISCUSSION AND IMPLICATION --- p.325Chapter 4.5.1 --- Feature Selection --- p.325Chapter 4.5.2 --- Mangrove Classification --- p.342Chapter 4.6 --- SUMMARY --- p.351Chapter CHAPTER 5 --- RESULTS AND DISCUSSION (II) - LEAF AREA INDEX MODELING --- p.353Chapter 5.1 --- INTRODUCTION --- p.353Chapter 5.2 --- DATA EXPLORATION --- p.353Chapter 5.2.1 --- Dependent Variable: Field measured LAI --- p.353Chapter 5.2.2 --- Independent Variables: Vegetation Index and texture measure --- p.355Chapter 5.2.3 --- Hyperspectral bands and LAI --- p.356Chapter 5.2.4 --- Normality testing --- p.359Chapter 5.2.5 --- Linearity testing --- p.363Chapter 5.2.6 --- Outliner detection --- p.365Chapter 5.3 --- SIMPLE LINEAR REGRESSION ANALYSIS --- p.366Chapter 5.3.1 --- LAI2000 Generalized method --- p.369Chapter 5.4 --- STEPWISE MULTIPLE REGRESSION ANALYSIS --- p.381Chapter 5.4.1 --- LAI2000 Generalized method --- p.384Chapter 5.5 --- DISCUSSION AND IMPLICATION --- p.391Chapter 5.5.1 --- LAI model comparison --- p.391Chapter 5.5.2 --- Species composition and LAI --- p.393Chapter 5.5.3 --- Hyperspectral Bands, Vegetation Indices and LAI --- p.397Chapter 5.5.4 --- Backscatter, texture measures and LAI --- p.407Chapter 5.5.5 --- Complementarity of Vegetation Index and Radar Parameters --- p.414Chapter 5.6 --- SUMMARY --- p.421Chapter CHAPTER 6 --- CONCLUSION --- p.423Chapter 6.1 --- SUMMARY OF THE STUDY --- p.423Chapter 6.2 --- LIMITATION OF THE STUDY --- p.427Chapter 6.3 --- RECOMMENDATION --- p.431Chapter REFERENCE --- p.434Chapter APPENDIX A --- GEOMETRIC CORRECTION OF HYPERSPECTRAL DATA --- p.473Chapter APPENDIX B --- SCRIPTS DERIVED FROM FEATURE SELECTION TOOLBOX (FST) FOR FEATURE SELECTION --- p.475Chapter APPENDIX C --- PREDICTED LAI(BON) AND LAI(2000) FROM SIMPLE LINEAR REGRESSION MODELS --- p.513Chapter APPENDIX D --- PREDICTED LAI(BON) AND LAI(2000) FROM MULTIPLE STEPWISE REGRESSION MODELS --- p.52