基于高光谱的晚疫病胁迫下马铃薯叶片的过氧化物酶活性检测

Detection of peroxidase activity in potato leaves under stress of late blight using hyperspectral imaging

针对马铃薯晚疫病难以实时无损检测的问题,提出了基于X-LW-PLS的模型,用以预测马铃薯晚疫病叶片高光谱信息与过氧化物酶(peroxidase,POD)活性之间的关联.为了降低光谱数据维度,提高模型运算速率,结合了连续投影算法SPA和载荷系数法X-LW选取特征波长来建立预测模型.测定不同染病时段(0,24,48,72,96 h)马铃薯叶片的高光谱信息和相对应的过氧化物酶POD活性值,利用ENVI软件提取样本的光谱反射特性曲线并结合多种化学计量学方法,建立马铃薯晚疫病叶片高光谱信息与过氧化物酶POD活性之间的关联预测模型.结果表明:基于全光谱信息的LS-SVM预测模型具有较好的预测效果,其校正集相关系数RC为0. 916,均方根误差RMSEC为19. 539 U·(g·min)^(-1),预测集相关系数RP为0. 932,均方根误差RMSEP为14. 966 U·(g·min)^(-1);而X-LW-PLS模型的预测效果最优,其RC为0. 870,RMSEC为37. 969 U·(g·min)^(-1),RP为0. 892,RMSEP为28. 922 U·(g·min)^(-1).利用高光谱技术来实现马铃薯晚疫病的实时无损检测是可行的.

To solve the difficulty of detecting potato late blight in real time,a prediction model was proposed based on X-LW-PLS model to predict the correlation between hyperspectral information of potato late blight leaves and POD activity of peroxidase.To reduce the dimension of spectral data and improve the operation speed of model,the prediction model was established by combining the continuous projection algorithm of SPA and the load coefficient method of X-LW to select the characteristic wavelengths.The hyperspectral information and the peroxidase(POD)activity of potato leaves at different infection periods of 0,24,48,72 and 96 h were measured.The spectral reflectance curves of samples were extracted by ENVI software and combined with various chemometrics methods to establish the association prediction model of hyperspectral information and peroxidase activity of potato late blight leaves.The results show that the LS-SVM prediction model based on the full spectrum information has good prediction effects with R C of 0.916,RMSE C of 19.539 U·(g·min)^-1 for calibration sets,R P of 0.932 and RMSE P of 14.966 U·(g·min)^-1 for prediction sets.The X-LW-PLS model has the best prediction effects with R C of 0.870,RMSE C of 37.969 U·(g·min)^-1,R P of 0.892 and RMSE P of 28.922 U·(g·min)^-1.It is feasible to use real-time hyperspectral technology to detect potato late blight.