Classification of wetland vegetation in the Yellow River Delta based on the fusion of hyperspectral data

Abstract

The Yellow River Delta wetlands are rich in vegetation species and the ecosystem is sensitive and fragile, so it is of great practical significance to accurately extract the information of the vegetation in the Yellow River Delta wetlands and analyze its spatial and temporal changes. Based on Sentinel-2A multispectral images and ZY1-02D hyperspectral images, this study addresses the problem of high spectral resolution and low spatial resolution of hyperspectral images, and adopts the Gram-Schmidt fusion method to fuse the two kinds of images, and uses the Random Forest Classification method to complete the classification of the typical vegetation of the Yellow River Delta region, and compares the classification accuracies of the three kinds of images. The study shows that: (1) Gram-Schmidt is more effective for image fusion, and the fusion of fused hyperspectral images based on the NIR band of Sentinel-2A images has the best fusion effect, which maximally retains the original spectral reflectance values, and at the same time improves the spatial resolution. (2) The classification accuracy of the fused hyperspectral images is high, with an overall classification accuracy of 85.62%, which is higher than that of single hyperspectral images and multispectral images.