马铃薯晚疫病叶片SPAD值的高光谱研究

对晚疫病害胁迫下马铃薯叶片中叶绿素含量(SPAD)及植被指数与高光谱特性进行了关联研究。首先,对接种晚疫病菌的马铃薯叶片进行连续观测7天,获取375~1 018nm波段范围内叶片的高光谱信息和SPAD信息;然后,应用ENVI软件提取高光谱数据中感兴趣区域的平均光谱信息,选择预处理效果最好的方法建立不同的预测模型,包括基于全光谱和连续投影算法(Successive Projection Algorithm,SPA)选择特征波长的偏最小二乘回归(Partial Least Squares Regression,PLSR)模型与最小二乘支持向量机(Least Squares Support Vector Machines,LS-SVM)模型,以及4种植被指数RENDVI、VOG1、VOG2、VOG3建立的简单经验估测模型。建模结果表示:全光谱的PLSR模型预测效果最好,而基于特征波长的LS-SVM模型和植被指数的二次多项式模型预测效果稍差,但大大简化了模型,为后续开发便携式仪器提供理论依据。研究表明:应用高光谱技术检测遭受晚疫病害的马铃薯叶片的SPAD及植被指数是可行的,为病害胁迫下作物的其他化学指标实时监测提供了新的路径。

SPAD Measurement of Potato Leaves Infected Late Blight Based on Hyper-spectral

The relation between the content of chlorophyll ( SPAD ) on the vegetation indexes and hyper-spectral feature in potato leaves under the stress of late blight was studied in this paper .Potato leaves that were inoculated with pathogen of potato late blight were observed for 7 continuous days , and the hyper-spectral data from 375 to 1018nm and the SPAD value were acquired every day .The average spectral data was extracted by analyzing the region of interest ( ROI) of the samples’ hyper-spectral data using the ENVI software ,choose the best pretreatment method to build different estimation models , including:a SPAD value estimation model based on partial least squares regression ( PLSR ) and least squares support vector machines ( LS-SVM) , a SPAD value estimation model based on successive projections algorithm combined with PLS and LS-SVM ( SPA-PLSR、SPA-LS-SVM ) ,4 kind of simple experience SPAD value estimation models .The results indicate that the optimal prediction performance is achieved by PLSR model in the whole wavelength region , LS-SVM model based on Characteristic wavelength and the quadratic polynomial model based on vegetation index have a slight poor prediction effect , but greatly simplify models , and provide theoretical basis for subsequent development of portable instrument .The research shows that it is feasible to detect SPAD of potato leaf suffered late blight and vegetation indexes using hyper-spectral , at the same time it provides a new path to implement real-time monitoring of other chemi-cal indexes under crop disease stress .