结合拉普拉斯特征映射的权重朴素贝叶斯高光谱分类算法

A Weighted Naive Bayes Hyperspectral Classification Algorithm Combined with Laplacian Eigen Mapping

高光谱遥感可以得到更精确与丰富的遥感信息,因此涵盖了各国家的航空、航天以及小范围的地面观测的多个层级与环节,在对地观测遥感领域占有不可取代的地位。但高光谱数据集往往非常庞大,且包含冗余信息,为后续处理带来了不便。该研究选用拉普拉斯特征映射对高光谱数据降维与特征提取,并提出了一种权重朴素贝叶斯分类算法。通过奖励权重的方法对经典朴素贝叶斯分类器进行了改进,利用公开数据对算法进行验证,判别地物信息准确率达到92.7%,相比于传统方法有了大幅度的提高。

Hyperspectral remote sensing,which plays an important role in the field of earth observation and remote sensing,could be used to obtain more accurate and rich remote sensing information,thus covering various levels and full links of the various countries′aerial,spaceflight and small range of ground observation.However,hyperspectral data sets are often very large and contain redundant information,which brings inconvenience to subsequent processing.In this study,Laplacian Eigen mapping was used to reduce the dimension and fulfil the feature extraction of hyperspectral data.Then a weighted naive Bayes classification algorithm was proposed,while the classic naive Bayes classifier was improved by the method of rewarding weight.The algorithm was verified by the open source data.Results indicated that the accuracy for the proposed method in identification of the object information reached to 92.7%,which was greatly improved compared with that for the traditional method.