基于冠层高光谱数据与马氏距离的马铃薯品种识别

Variety identification of potatoes based on canopy hyperspectral data and Mahalanobis distance

为丰富高光谱数据在精细农业中的应用,本研究基于冠层光谱数据进行不同马铃薯品种区分研究。利用田间实测的6-8月的马铃薯原始光谱数据以及经过一阶微分、对数一阶微分、包络线去除处理后的光谱,采用马氏距离法选择3种马铃薯光谱差异显著波段,再利用逐步判别法检验波段识别精度。结果表明,7月份经过对数一阶微分变换选取的特征波段识别精度最高,达87. 7%。不同生育期内,多种预处理方法下的光谱识别能力有差异。6月份包络线去除法的识别精度最高,7月份对数一阶微分处理下的识别精度最高,而8月份原始光谱的识别精度最高。提取的特征波段多位于红光及近红外波段。研究结果表明基于高光谱数据,借助马氏距离与逐步判别法可以区分马铃薯品种。

In order to enrich the application of hyper-spectral data in precision agriculture,this study attempts to distinguish potato varieties based on canopy spectral data.Based on the field measured data of the original spectra of potato from June to August,and the spectra processed by the first order differential,logarithmic first order differential and continuous removal,the method of Mahalanobis distance and stepwise discriminant were used to select characteristic bands for variety identification.The result showed that feature bands extracted from the spectra of the logarithmic first order differential transformation in July had the highest recognition accuracy of 87.7%.The recognition ability of different preprocessing spectra was different in growth stages.In June,the accuracy of continuous removal was the highest,and the accuracy of logarithmic first order differential treatment in July was the highest,while the original spectrum in August had the highest accuracy.The extracted characteristic bands were mostly in red and near infrared bands.This study shows that based on hyperspectral data,potato varieties can be distinguished by Mahalanobis distance and stepwise discriminant method.