Research on EnMap-BOX Land Use Classification Method Based on Multi-feature Combination
In order to study the application effect of multi-feature combination in land use classification of high-resolution satellite im-ages in China. In this paper, GF6-WFV multi-spectral image data was used to build a multi-feature combination based on spectral band, vegetation index and texture features, and ReliefF algorithm was used for feature optimization to obtain the optimal feature set with less information redundancy. The optimal classification model was obtained by combining the penalty parameter C and kernel function coefficient g in the improved SVM algorithm of EnMap-BOX toolkit, and the study area was classified. The results show that: 1) Feature selection can effectively reduce the information redundancy of multi-feature sets. 2) The improved SVM algorithm model based on feature selection can obtain a high land use classification accuracy, with the overall accuracy reaching 82.89％ and the Kap-pa coefficient reaching 0.78, which can provide a method with high application value for land use classification.
ReliefFEnMap-BOXimproved SVM algorithmland use classification
万方数据
维普
