|Title:||由植被高解析反射光譜模式化稻株之生長||Other Titles:||Using Hyperspectral Reflectance Data to Modeling Rice Growth||Authors:||楊純明
|Keywords:||高解析反射比光譜;水稻生長模式化;直線複迴歸;光譜特徵;光譜指數;Hyperspectral reflectance spectrum;Rice growth modeling;Multiple linear regression;Spectral characteristic;Spectral index;Rice Precision Farming System||Issue Date:||Sep-2003||Publisher:||農業試驗所||Start page/Pages:||13-24||Source:||農業試驗所特刊第105號||Conference:||水稻精準農業體系||Abstract:||
本文研究旨在探討近地面量測之水稻植被高解析反射光譜之季節變化，並試以利用不同方法來篩檢估測水稻生長的光譜特徵並建立光譜遙測模式。高解析光譜係以田間攜帶式高解析輻射光譜儀偵測，生長性狀則於量測光譜時取樣調查，試驗期間為2000-2002 年之一二期稻作，計有四期作。篩檢之光譜特徵為可見光之綠光波段峰點（G REEN）、紅光波段之谷點（RED）及近紅外光之波段頂點（NIR)，而光譜指數係由此三項動態特徵之反射比計算，包括RRED/RNIR ratio、RGREEB/RNIR ratio、RRED/RGEEN ratio及NDVI (normalized difference vegetation index，標準差植被指數或稱正規差植生指數）。試驗發現稻株之生長性狀於抽穗前後達到最高點，再隨著成熟老化而下降。據此，將水稻生育全期以抽穗為分割點劃分為抽穗前期（pre-heading phase ）及抽穗後期（post-heading phase ) ，可提高光譜特徵及光譜指數與生長性狀間之相關性。又由多元直線複迴歸（MLR）分析，可經由對決定係數（coefficient of determination , R2的要求，來選取光譜中合適的光譜特徵波段數目以建立多元直線複迴歸模式估測稻株之生長。根據研究結果，多元直線複迴歸模式確實提供了選取光譜特徵的彈性，同時也改進了對稻株生長變異的估測。
Experiments were conducted to study the seasonal changes of rice canopy reflectance spectra from near ground platform and to modeling rice (Oryza sativa L. cv. Tainung 67) growth from the related spectral characteristics and spectral indices. Reflectance spectra were taken by a field-portable spectroradiometer and growth parameters were measured after spectral measurements regularly during the first and the second cropping seasons in 2000-2002. Spectral indices of RRED/RNIR ratio, RGREEN/RNIR ratio, RRED/RGREEN ratio and normalized difference vegetation index (NDVI) were calculated from dynamic spectral characteristics GREEN, RED and NIR. Seasonal changes of the spectral characteristics, spectral indices, and growth parameters were either curvilinear or linear in both crops. It was found that growth parameters reached the maximum near heading when the vegetative growth climbed to the climax, and decreased after this stage when the plants grown toward maturity. Thus, modeling between spectral characteristics/indices and growth parameters may be improved by separating growing period into the pre-heading and post-heading phases. By the multiple linear regression (MLR) analysis, the best-2 or more variables MLR models were determined, which can be defined with an user-dependent coefficient of determination (R2) for numbers of wavebands added to the models. The MLR models provided flexibility in choosing the individual narrow bands and improved the relationships to estimate plant growth.
|Appears in Collections:||作物組|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.