摘要
为量化判别烟叶成熟度,使用ASD Fieldspec FR2500光谱仪测定和分析了5个不同成熟度烤烟鲜烟叶的可见/近红外(350-1650 nm)光谱.结果表明:不同成熟度鲜烟叶在可见光503-651 nm之间光谱差异极其显著.通过逐步回归分析筛选出3个主要预测因子514、629、650 nm的反射率的典则判别分析模型.判别模型的准确率检验结果,训练样本为98%,验证样本为97%.基于高光谱的烟叶成熟度判别模型具有较好的稳健性.使用光谱技术进行鲜烟叶成熟度判定是可行的.
Spectra of leaves of 5 maturity grades in visible/near-infrared （350 - 1650 nm） wavelengths were measured and analyzed to identify grades digitally at flue-cured virginia tobacco harvesting time. The result showed that the reflectance spectra distinguished significantly from grades each other in 503 -651 nm. Fisher＇s linear discriminant models with reflectance of 514 nm, 629 nm and 650 nm have been developed by stepwise regression and have been further tested and calibrated. The accuracies of training set and calibration set were 98% and 97% , respectively. Maturity grades were robustly estimated with discriminant model based on hyper- spectrum. It is feasible to identify leaf maturity grade accurately with spectral technology.
出处
《福建农林大学学报（自然科学版）》
CSCD
北大核心
2008年第6期565-569,共5页
Journal of Fujian Agriculture and Forestry University:Natural Science Edition
基金
国家"863"计划(2004AA115190)
昆明市科技局创新基金(cj2006001)资助项目
关键词
烤烟
成熟度
光谱特征
判别分析
flue-cured virginia tobacco
maturity grades
spectral characteristic
discriminant analysis
