Abstract: Aiming at the problem that the fusion algorithm has a significant impact on the accuracy of image classification, this paper chooses the coastal wetland of Lianyungang Liezi Estuary coastal zone as the research area, and GF-1 satellite image is used as data source. Three fusion algorithms, Gram-Schmidt, PCA and Brovey are used to fuse images. After that, on the eCognition software platform, based on the object-oriented multi-scale segmentation technology, the image is classified by random forest algorithm, and evaluate the accuracy of the classification result. Experimental results show that the image quality after Gram-Schmidt fusion algorithm is the best, and the classification accuracy of Gram-Schmidt algorithm is the highest; Brovey fusion algorithm has good spectral fidelity for vegetation and water body,after changing the band combination, the classification accuracy of Brovey algorithm is obviously improved; the accuracy of PCA algorithm is the lowest among the three fusion algorithms.

Key words: image fusion, land use classification, object-oriented, random forest, accuracy evaluation