基于非成像高光谱的云烟97号叶片钾含量估算研究

摘要

To quickly detect the potassium content per unit weight in tobacco leaves nondestructively, the leaf spectrums of Yunyan 97 as the main cultivar in Damogu Town, Qujing City with different growth vigor were collected by the ASD Field Spec 3 portable ground object spectrometer. The potassium contents of corresponding leaves were determined in lab. With the spectral analysis technology, the variables of spectral characteristics of tobacco leaves were extracted to analysis the correlations between the potassium content and original spectrum, the first derivative of the spectrum and hyperspectral characteristics. By the method of multiple stepwise regression, the potassium content estimation model was established based on the single variable of spectral characteristics, and the accuracy of the model was tested. The results showed that the correlation coefficient between K content and original spectral reflectance was the largest at 350 nm. The correlation between K and the first derivative of the spectrum was negatively extremely significant at 1 089 nm with the value of-0. 661, and positively extremely significant at 2 297 nm with the value of 0. 710. The correlation coefficient between K and green peak area (SDg) was the largest. To sum up, the regression model based on the first derivative of spectral reflectance had higher accuracy and better estimation effects of the potassium content in tobacco leaves.%为无损快速检测基于单位重量的烟叶钾元素含量,应用ASD Field Spec 3便携式地物光谱仪采集了曲靖市大莫古镇主栽品种云烟97号不同长势的烟草叶片光谱,测定了相应烟叶的钾元素含量;利用光谱分析技术提取烟草叶片光谱特征变量,并分析了烤烟叶片钾元素含量与原始光谱、光谱一阶微分及高光谱特征变量间的相关性;采用多元逐步回归分析的方法,建立了基于光谱特征单变量估测钾含量模型并进行该模型的精度检验.结果 表明,钾元素(K)与原始光谱反射率的相关系数在350 nm波长处最大;钾元素(K)与光谱一阶微分在1 089 nm波长处极显著负相关(相关系数为-0.661),在2 297 nm处极显著正相关(相关系数为0.710);钾元素与高光谱特征变量绿边面积(SDg)的相关系数最大.综合来看,基于叶片光谱反射率一阶微分的钾元素回归模型对烟草叶片钾含量的估测精度较高,估测效果较好.