基于遗传算法近红外光谱检测土壤养分的研究

Study on Soil Nutrients by Using Near-Infrared Spectroscopy Based on the GA method

利用可见/短波近红外光谱(Vis/SW-NIRS)分析测量土壤速效氮(N)和速效钾(K)。探讨遗传算法在分析测量土壤养分上的应用,根据优化结果采用最小二乘支持向量机(LS-SVM)方法建立校正模型。结果表明,LS-SVM模型优于PLS模型;GA-LS-SVM模型预测速效氮的精度较高。基于遗传算法可见/短波近红外光谱利用LS-SVM建模,可以作为一种土壤理化性质的测定方法。

Visible infrared spectroscopy （Vis/SW-NIRS） was investigated to study the measurement accuracy of soil available nitrogen（N） and available potassium （K）. Correction model was established using LS-SVM according to the optimal results in order ot explore the application of genetic algorithm in measuring soil nutrients. The results indicated that, LS-SVM models outperformed PLS models; Compared with available K, forecasting precision of GA-LS-SVM models was more preferable for available N. Visible and short wave-near infrared spectroscopycombined with LS-SVM based on GA could be utilized as a method for the determination of soil properties.