摘要
为了分析不同质地土壤的近红外光谱特性,建立合适的土壤质地分类预测模型。研究以沙土、壤土和黏土3种不同类型土壤作为研究对象,采集了山西省内3个地区的土壤样本共156个,获取其近红外光谱数据,采用支持向量机(SVM)在1 001~2 500 nm波段内对不同质地土壤的吸光度值进行建模预测。结果表明,3种质地土壤具有不同的光谱反射特性;利用支持向量机建立的土壤分类预测模型,其测试集的预测正确率达到91.67%,说明SVM在土壤分类应用中的效果较好,可以利用SVM模型进行土壤属性预测。
To analyze the near infrared spectral characteristics of different texture soils, a suitable prediction model was established.In this study, 3 types of soil were studied, including sand, loam and clay, the 156 soil samples were collected from 3 regions in Shanxi province and the data of the near infrared spectra were obtained. Support vector machine(SVM)was used to model and predict the absorbance value of different texture soils in the 1 001-2 500 nm band. The results showed that three kinds of soil texture had different spectral reflectance characteristics. Soil classification prediction model by using support vector machine, predict accuracy of the test set reached 91.67%, indicating good effect in the application of SVM in soil classification, soil properties can be predicted by the SVM model.
出处
《山西农业科学》
2017年第10期1643-1645,1654,共4页
Journal of Shanxi Agricultural Sciences
基金
国家自然科学基金项目(41201294)
山西省科技攻关项目(20130313010-6)
关键词
质地
近红外
SVM
spectral characteristics
different texture soils
prediction model
established
soil
regions
Shanxi province