土壤碱解氮含量可见/近红外光谱预测模型优化

可见/近红外光谱技术是土壤成分检测的有效工具。波长筛选对可见/近红外模型土壤属性的预测精度有重要影响。以宁夏吴忠地区75个水稻土样为研究对象，利用可见/近红外光谱技术采集土壤样品光谱，采用SPXY （Sample set partitioning based on joint X-Y distance）方法选取了校正集和预测集样本，比较了分别采用Savitzky Golay平滑（SG smoothing）、多元散射校正（Multiple scatter correction，MSC）、标准正态变量变换（Standard normal variate，SNV）3种预处理方法对光谱数据处理后建立土壤碱解氮偏最小二乘法模型和原始光谱数据建模的效果。在此基础上，分别采用遗传算法（Genetic gorithms，GA）、连续投影算法（Successive projections algorithm，SPA）、竞争性自适应重加权算法（Competitive adaptive reweighted Sampling，CARS）、随机蛙跳（Random frog，RF）进行波长筛选，最后应用偏最小二乘法建立基于不同波长筛选方法的土壤碱解氮含量预测模型。研究表明，由于仪器性能稳定，样品的颗粒度比较小和均匀，本次实验原始光谱数据建模效果最好；各种波长筛选方法均可有效减少参与建模的波长数，且连续投影算法优于全谱建模，所选波长数仅为全谱波长数的1%，其预测决定系数（R²）、预测均方根误差和相对分析误差值分别为0.726，3.616，1.906。这表明连续投影算法可以有效筛选水稻土碱解氮敏感波段，为土壤碱解氮传感器开发提供技术支持。

Optimization for Vis/NIRS Prediction Model of Soil Available Nitrogen Content

Visible/near infrared spectroscopy(Vis/NIRS) is an effective tool for soil component detection. Wavelength selection plays an important role in predicting the soil properties of the visible/near infrared(Vis/NIR) model. Taking the 75 paddy soil samples of Ningxia Wuzhong area as the research object, the Vis/NIR spectra of soil samples were collected. The sample set partitioning based on joint X-Y distance(SPXY) method was used to divide calibration set and prediction set samples. Three different preprocessing methods, namely, Savitzky Golay smoothing(SG smoothing), multiplicative scatter correction(MSC), and standard normal variate(SNV), were applied to pretreat the spectral data. Then, three soil available nitrogen content prediction models were established by partial least squares(PLS) regression. The consequences of the three models and the model using raw spectral data were compared. On this basis, the genetic algorithm(GA), successive projections algorithm(SPA), competitive adaptive reweighted algorithm(CARS) and Random Frog(RF) were used to select key wavelength variables. Finally, the soil available nitrogen content prediction models were established by partial least squares regression based on different wavelength selection methods. Research shows that, due to the stability of the instrument, relatively small and uniform of the sample particle size, raw spectral data is achieved the best consequence; various wavelength selection methods can effectively reduce the number of wavelengths in the model, and the continuous projection algorithm is better than full spectrum model, the selected wavelength number is only 1% of the number of full spectrum wavelength and its prediction coefficient of determination(R²), root mean square prediction error and relative error analysis values were 0.726, 3.616 and 1.906. The research shows that wavelength selection method by SPA could predict the available nitrogen content in the paddy soil, provide a reference for development of soil available nitrogen sensor.