提出了一种无损、快速、成本低的土壤分类方法，选取松嫩平原4种典型土壤（黑土、黑钙土、风砂土和草甸土）耕层（0-20 cm）土样的实验室反射光谱数据作为研究对象，采用重采样、包络线消除法处理光谱数据，提取反映反射光谱特征的光谱特征参数，利用K均值聚类（K-means clustering）和决策树（decision tree）分别进行聚类分析和分类模型构建，实现土壤的快速分类。结果表明，利用表层土壤反射光谱特征参数构建的决策树分类模型可以对研究区土壤进行分类。研究成果有望加快土壤制图，为土壤理化性质的时空变化研究提供技术支持。

Soil taxonomy plays a significant role in soil remote sensing. Soil spectral reflectance is the comprehensive representation of soil's physical and chemical parameters. The study of soil spectral reflectance features is the physical basis for soil remote sensing, and it provides new ideas and methods for soil classification.
To quickly classify soil based on topsoil reflectance spectral characteristics and provide an effective method, the room spectral reflectance in the visible and near-infrared region (400-2500 nm) of 148 soil samples, including black, chernozem, blown, and meadow soils,were collected from Songnen plain, which is located in Heilongjiang province. Given that the high-frequency noise of reflectance spectrum is relatively strong in the range of 400 nm to 430 nm and 2450 nm to 2500 nm, we chose the visible and near-infrared region of 430 nm to 2450 nm.
The spectral reflectance of soil samples were measured using ASDFieldSpec 3 in the laboratory. Resampling and continuum removal techniques were used to process spectral data and extract the spectral characteristic parameters (i.e., the absorption positions of the spectral curve, the vale's area, the slope of the spectral curve, the distance between adjacent absorption positions, the depth of the vale, and the width of the vale), respectively. When the K-means clustering results based on spectral reflectance were compared with the K-means clustering results based on the spectral characteristic parameters, the spectral characteristic parameters were found to be more suitable for soil classification. Finally, the spectral characteristic parameters were used to constructsoil classification model that is based on the decision tree. The classification accuracy of black, chernozem, blown, and meadow soilsare 97.22%, 94.2%, 85.29% and 55.56%, respectively.These results were obtained by using the decision tree model. The most effective spectral characteristic parameters include the second absorption positions of the spectral curve, the first vale's area, the first two vales' area, and the slope of the spectral curve at 500 nm to 600 nm and 1340 nm to 1360 nm. Meadow soilis often distributed in the lower area, and the spectral curve of the topsoil of meadow soil is similar to its adjacent soil, which is regarded asa trend toward its adjacent soil. The spectral characteristic parameters that were extracted could be used to study the soil classification, and the decision tree model that is based on the spectral characteristic parameters of topsoil reflectance has achieved excellent results. This paper provides a convenient, rapid, and nondestructive approach for soil classification, which helps in soil mapping.