基于主成分分析和BP神经网络的土壤养分近红外光谱检测

基于近红外光谱技术的土壤养分快速、无损检测,有利于精细施肥决策。在一黄豆田采用7 m×7 m的栅格采集54个土样,测定其土壤有机质、速效氮、有效磷、有效钾,并使用FieldSpec 3光谱仪测定土样的近红外漫反射光谱。将54个样本随机分成预测集与验证集,其中预测集40个,验证集14个。通过平滑预处理后,利用主成分分析法(PCA)提取原始光谱8个主成分。然后以8个主成分为输入,分别以所测土壤养分作为输出,建立土壤有机质、速效氮、有效磷、有效钾的预测模型,最后对14个验证样本进行预测。结果表明,在小尺度采样的情况下进行光谱分析,采用主成分分析和人工神经网络相结合的方法建立土壤有机质预测模型,其测量值与预测值的相关性较高,相关度为0.796 2,相对误差较小,其平均值为1.88%,表明该方法预测土壤有机质含量是可行的。但对土壤速效氮、有效磷和有效钾含量的预测并不理想,还有待进一步研究。

Near-Infrared Determination of Soil Nutrients Based on Principal component Analysis and BP Neural Network

In order to make decision for precision fertilization,based on the near-infrared spectroscopy technology,the detection method of the soil nutrients(organic matter and available N,P,K) have been analyzed.In a soybean field,54 samples by 7m×7m have been collected using the DGPS receiver positioning.The soil organic matter,available nitrogen,phosphorus and potassium content have been determined,and the near-infrared diffuse reflectance spectrum of the soil samples have been obtained using FieldSpec 3 spectrometer.Fifty four samples were randomly divided into the prediction set and the validation set with 40 and 14 samples respectively.After smoothing,the eight principal components of original spectra were extracted by using principal component analysis(PCA).And prediction model of soil organic matter,available nitrogen,phosphorus and potassium were respectively established with the eight principal component as input and soil nutrients by measured as the output,and the last of the 14 validation samples were predicted.The results show that the soil organic matter prediction model has been set up using principal component analysis and artificial neural network,which the correlation coefficient between the prediction value and measurement value is 0.7962,and the relative error of the neural network prediction is smaller and its mean value is 1.88%,the method of soil organic matter content prediction is feasible.