Remote sensing images have the characteristics of multiple data sources and complex data. How to integrate remote sensing image information more efficiently has always been the focus of research. In this paper, Changchun City, Jilin Province, China was selected as the experimental area, Sentinel-1 and Sentinel-2 images were used as experimental data, and a fusion method of SAR image and multispectral image using texture feature information was proposed. First, perform HIS transformation on the multi-spectral image to obtain the intensity image. After that, wavelet transform was used to extract the high-frequency and low-frequency detail components of the intensity image. At the same time, the principal component analysis method and the deep learning network VGG-19 were used to extract the texture features of the SAR image. The SAR texture image was used to enhance the high-frequency detail component of the intensity image, and combined with the original low-frequency detail component to perform inverse wavelet transform, then a new intensity image was obtained. Finally, the modulated intensity image was used to replace the original intensity image, and the inverse transformation (I-HIS) was performed to obtain an enhanced image fused from the multispectral image and the SAR image. Compared with the original image, the detailed features and boundary distinction were significantly improved. The fusion image was input into the support vector machine for feature classification, and the comprehensive classification accuracy reached 94.74%, which was 3.5% higher than the classification accuracy of the unfused image.
|