基于近红外光谱岭回归的甜椒氮素检测试验

Experiment on sweet pepper nitrogen detection based on near-infrared reflectivity spectral ridge regression

针对设施作物营养水平无损检测技术,着重描述了基于近红外光谱数据岭回归分析的甜椒氮素检测试验研究过程.利用近红外反射光谱成像技术对目标作物进行叶片尺度的光谱图像采集,应用计算机图像分析软件进行光谱数字图像处理、提取光谱数据,经过统计分析对数据完成筛选作为变量,结合化学分析试验结果建立作物营养检测模型,检验模型得出结论.为了解决自变量间存在的多重共线性造成模型难以建立的问题,在数据处理阶段,采用了在农业探测领域内并不多见的岭回归分析方法,利用其特殊的有偏估计算法,拟合建立回归方程.同时,由于岭回归分析可以用于进一步筛选特征波段,最终得到的是基于三特征波段近红外光谱反射率数据的甜椒叶片氮营养检测模型.经过模型检验,模型的调整R2为0.843,RMSE为0.105.

According to nondestructive nitrogen content detection techniques for greenhouse cultivated crops,an experimental process for sweet pepper nitrogen detection was described based on ridge regression analysis of near-infrared reflectivity( NIR) spectral data. Firstly,the targeted crop spectral images in leaf scale were captured by NIR reflectance spectral imaging technique; secondly,a spectral digital image processing was launched by using image analysis software to extract spectral data;thirdly,variables were screened by statistical analysis on the data; finally,a crop nutrition detection model was established by completing agricultural chemical analysis. In the data processing phase,ridge regression,which is a kind of biased estimation method,was adopted to obtain the regression equation.In the ridge regression analysis, characteristic bands were screened further, then, the nitrogennutrition detection model of sweet pepper leaf was gotten based on NIR spectra data of three characteristic bands. The adjusted R2 in the model is 0.843,the RMSE is 0.105.