中国烟草学报
中國煙草學報
중국연초학보
ACTA TABACARIA SINICA
2014年
1期
54-60
,共7页
邢雪霞%刘国顺%贾方方%孙榅淑%贾春雷
邢雪霞%劉國順%賈方方%孫榅淑%賈春雷
형설하%류국순%가방방%손온숙%가춘뢰
烤烟%高光谱%色素含量%模型
烤煙%高光譜%色素含量%模型
고연%고광보%색소함량%모형
tobacco%hyperspectral data%leaf pigment content%model
研究烤烟叶片的高光谱曲线特征,探索建立烤烟色素含量的高光谱预测模型,以促进高光谱技术在现代烟草农业中的发展。采用大田试验,分析了不同光质条件下,烤烟叶片光谱的特征。利用相关分析方法,确定了21个光谱参数与色素含量的相关性,并建立了叶片色素含量的高光谱线性与非线性模拟方程。不同光质处理下,烟叶叶片光谱曲线相似,在可见光与近红外短波区域差异比较明显,而在近红外长波区域基本没有差异。光谱参数G_NDVI和TCARI分别与叶绿素、类胡萝卜素含量之间有较好的相关性,并建立了预测模型。经精度检验结果显示,模型能较好的预测烤烟色素含量。光谱参数G_NDVI和TCARI能有效检测烤烟色素含量,为高光谱技术在不同生态区域的应用提供理论依据。
研究烤煙葉片的高光譜麯線特徵,探索建立烤煙色素含量的高光譜預測模型,以促進高光譜技術在現代煙草農業中的髮展。採用大田試驗,分析瞭不同光質條件下,烤煙葉片光譜的特徵。利用相關分析方法,確定瞭21箇光譜參數與色素含量的相關性,併建立瞭葉片色素含量的高光譜線性與非線性模擬方程。不同光質處理下,煙葉葉片光譜麯線相似,在可見光與近紅外短波區域差異比較明顯,而在近紅外長波區域基本沒有差異。光譜參數G_NDVI和TCARI分彆與葉綠素、類鬍蘿蔔素含量之間有較好的相關性,併建立瞭預測模型。經精度檢驗結果顯示,模型能較好的預測烤煙色素含量。光譜參數G_NDVI和TCARI能有效檢測烤煙色素含量,為高光譜技術在不同生態區域的應用提供理論依據。
연구고연협편적고광보곡선특정,탐색건립고연색소함량적고광보예측모형,이촉진고광보기술재현대연초농업중적발전。채용대전시험,분석료불동광질조건하,고연협편광보적특정。이용상관분석방법,학정료21개광보삼수여색소함량적상관성,병건립료협편색소함량적고광보선성여비선성모의방정。불동광질처리하,연협협편광보곡선상사,재가견광여근홍외단파구역차이비교명현,이재근홍외장파구역기본몰유차이。광보삼수G_NDVI화TCARI분별여협록소、류호라복소함량지간유교호적상관성,병건립료예측모형。경정도검험결과현시,모형능교호적예측고연색소함량。광보삼수G_NDVI화TCARI능유효검측고연색소함량,위고광보기술재불동생태구역적응용제공이론의거。
Hyperspectral characteristics of tobacco leaves were analyzed with the objective to build hyperspectral monitoring models for pigment content in tobacco leaves. The development of hyperspectral techniques in modern tobacco agriculture was discussed. Hyperspectral characteristics of tobacco leaf with different light constitution were investigated. Relationship between 21 hyperspectral parameters and pigment content was analyzed by using relevance analysis. Hyperspectral parameter-based linear and nonlinear monitoring models for tobacco leaf pigment content estimation were built. Hyperspectral curves under different light quality were similar. Difference in visible and near-infrared shortwave regional was signiifcant, while there is little difference in the near-infrared wavelength. Spectral indices of G_NDVI and TCARI showed a close correlation with chlorophyll and carotenoid content, and their monitoring models were set up accordingly. These models were proved to be reliable for monitoring content of leaf pigment monitoring by the index of root mean square deviation (RMSE) and relative error (RE%). Results showed the availability of G_NDVI and TCARI in monitoring leaf pigment content , thus providing a theoretical basis for the application of hyperspectral technology in different ecological areas.