基于高光谱技术的土壤镉污染分级评价研究  

Research of soil cadmium pollution grading evaluation based on hyperspectral technology

作　　者：刘彦姝[1] 潘勇[2,3] 

机构地区：[1]湖南大众传媒职业技术学院,湖南长沙410100 [2]中南大学地球科学与信息物理学院,湖南长沙410083 [3]长沙师范高等专科学校,湖南长沙410083

出　　处：《生态环境学报》2012年第7期1361-1365,共5页Ecology and Environmental Sciences

基　　金：国家"973"计划前期研究专项(2007CB416608)

摘　　要：提出一种利用高光谱技术进行土壤镉污染分级评价的方法。以FieldSpec 3地物光谱仪采集厂矿区土壤光谱反射率175份,随机分成校正集(135份)和检验集(40份)。光谱经小波去噪和标准归一化(SNV)处理后,以主成分分析法(PCA)降维。将降维所得的前5个主成分数据为输入变量,分别采用Fisher线性判别、Byes逐步判别、模糊模式识别、BP-ANN判别以及SVM 5种方法建立了土壤镉污染分级评价模型,并利用40个未知样对模型进行检验。结果表明:Fisher线性判别准确率为77.5%,Byes逐步判别与模糊模式识别预测为80.0%,BP-ANN模型预测精度为82.5%,SVM模型预测精度最高,达85.0%。说明采用高光谱技术进行土壤镉污染分级评价是可行,其中SVM是建模的优选算法。A new method was put forward to assess the cadmium pollution of soil by hyperspectra technology.175 soil samples were collected using a FieldSpec 3 spectrometer in mine and factory area.All samples were divided randomly into 2 groups,one group with 135 samples used as calibration set,and another with 40 samples used as prediction set.The samples data were pretreated with the methods of wavelet denoising and standard normalization variate（SNV）,and then analyzed by principal component analysis（PCA）.The top 5 principal components of PCA were used as the new variables,and modeling by fisher linear discrimination,Bayes multi-types stepwise discrimination,fuzzy pattern recognition,back-propagation artificial neural network（ANN-BP） and support vector machine（SVM） algorithms.Then,the 40 unknown samples in the prediction set were predicted,and the result showed that the discriminating accuracy rate was 77.5% with the methods of fisher linear discrimination,80.0% with the method of Bayes multi-types stepwise discrimination and fuzzy pattern recognition,82.5% with the method of BP-ANN model,and 85.0% with the method of SVM model.Therefore,the feasibility of assessment the cadmium pollution of soil in rapid and non-invasive way by hyperspectra technology was proved,and PCA combined with SVM was confirmed as a preference method.

关 键 词：高光谱 镉污染 土壤 判别 支持向量机 

分 类 号：TP722[自动化与计算机技术—检测技术与自动化装置] S714.5[自动化与计算机技术—控制科学与工程]