论文部分内容阅读
为了提供果树精准肥水管理参考数据,进行了果树叶片SPAD值近红外光谱无损检测研究。采用反射方式,采集100片赣南脐橙叶片的可见近红外光谱;利用移动窗口偏最小二乘法结合遗传算法、连续投影算法筛选光谱变量,并建立偏最小二乘回归校正模型。采用移动窗口偏最小二乘法和连续投影算法组合筛选的39个光谱变量建立的校正模型预测结果最优,模型预测相关系数为0.898,模型预测SPAD值均方根误差为2.116。试验表明,应用可见近红外反射光谱技术结合化学计量学算法进行赣南脐橙叶片SPAD值无损检测是可行的。
In order to provide reference data of fruit tree precision fertilizer and water management, the nondestructive testing of fruit leaf SPAD value near infrared spectroscopy was carried out. The visible near infrared spectra of 100 Gannan navel orange leaves were collected by reflection method. The moving window partial least squares method combined with genetic algorithm and continuous projection algorithm were used to screen the spectral variables and the partial least squares regression model was established. The calibration model established by the combination of moving window partial least square method and continuous projection algorithm was the best. The model predictive correlation coefficient was 0.898. The model predicted SPAD value was 2.116. Experiments show that the application of visible near-infrared reflectance spectroscopy combined with stoichiometric calculation method of Ganoderma lucidum leaves SPAD nondestructive testing is feasible.