为实现不同种类土壤的快速分类鉴别, 实验研究了基于激光诱导击穿光谱技术的土壤快速分类方法。 由于不同类型的土壤在元素组成上会存在较大差异, 所以利用激光诱导击穿光谱技术进行土壤分类具有可行性。 不同土壤在相同实验条件下产生的等离子体温度会存在较大差异, 可以作为分类的重要依据, 所选择的7类土壤中, 赤红壤的等离子体温度最高。 选取土壤中6种常量元素Si, Fe, Al, Mg, Ca和Ti的光谱强度作为分类指标, 利用主成分分析（principal component analysis, PCA）对7种土类的25个样品进行了分类, 其中砖红壤和赤红壤分类出现了交叠, 而不同高山草甸土样品之间元素差异较大, 并没有实现较好的聚类。 利用反向传播神经网络(back-propagation artificial neural network)结合土壤的LIBS光谱对土壤进行了分类, 分类结果与PCA结果相近, 赤红壤与砖红壤出现了识别错误。 当用PCA分析获得三个主成分值作为BP神经网络的输入量时, 获得了较好的分类结果, 因为简化了输入量, 降低了BP神经网络的误差, 此时只有一个高山草甸土被识别成褐土, 而高山草甸土的等离子体温度显著低于褐土, 所以结合不同土壤类型的等离子体温度差异, 能够实现不同土壤的分类识别。 实验证明激光诱导击穿光谱技术可以应用于土壤分类, 为土壤普查和合理利用提高了一种新的技术。

Soil classification is an important research content in soil science field. It is the basis of soil survey and resource evaluation which is important to agricultural production. There are many soil classification standards all over the world. China has two kinds of classifications including soil genetic classification and soil system classification. There are great differences between different types of soil elements, so it is feasible for soil classification to use laser induced breakdown spectroscopy. Laser induced breakdown spectroscopy (LIBS) is a new element analysis technology which uses a laser pulse with high energy density to ablate samples. LIBS has been used in many fields including environmental protection and industrial production control. It can directly reflect the difference of element content in different soils. The experimental setup including an Nd: YAG laser, a spectrometer, a computer and a rotating platform. In the experiment 7 kinds of soil (red soil, brick red soil, lateritic red soil, paddy soil, cinnamon, alluvial soil and alpine meadow soil) including 25 samples were used. All soil samples were grinded and sieved before the experiment.Under the same experimental condition, the temperatures of the plasma created by the laser pulses on the surface of the different soil samples have great differences. The lateritic red soil had the highest temperature, and the alpine meadow soil had the lowest. But it was not enough to form the basis for classification. Therefore six constant elements including Si, Fe, Al, Mg, Ca and Ti were selected and their spectral line intensity were treated as classification index. Principal component analysis (PCA) was used to simplify the classification process. The PCA method could simplify the 6 indexes to few independent indexes which could also reflect the spectral information of the 6 elements. The original spectral data was processed by Matlab. The process consisted of spectral background removal, characteristic spectrum identify and extraction. The classification results showed a three--dimensional figure. Except alpine meadow soil which varied in element concentrations 6 kinds of soils achieved good classification. The brick red soil and lateritic red soil varied in PC1, but their PC2s and PC3s were the same. The two kinds of soil overlapped with each other and they couldn’t be separated. Back-propagation artificial network was also used to achieve soil classification. The classification results were the same with the PCA. Some brick red soil and lateritic red soil samples were identified inaccurately. When the PC1, 2, 3 were used as the input of the BP-neural network, the classification had much better accuracy because less input improved the performance of the BP-neural network. Only one alpine meadow soil sample was identified to cinnamon soil. When the plasma temperature was also taken into account, all the soil samples could be distinguished. The results showed that LIBS could be used to classify soils based on their element content differences. The PCA, soil plasma temperature and BP-neural network were useful tools to achieve soil classification. The LIBS provides a useful tool for general detailed soil survey and rational utilization of soil.