Application of region selection network-based monitoring in blurring degradation of high-resolution optical remote sensing images
Jianghao Liu, Xin Mu, Shuhai Yu, Yunhe Liu, Chong Wang, Kai Zhou · 2025
In this paper, an automatic extraction method of fuzzy kernel for remote sensing images is proposed for the core evaluation index of remote sensing imaging clarity. In this paper, the deep learning based two-stage VIT classification network TIMCT (Two-stage Improved Multi-Scale Classified Transformer) is firstly trained, by using the dataset with known remote sensing image feature types to train the network and classify the images, these output scores obtained through training with high options, i.e., the network selects the region results. Then the remote sensing image is screened in two stages using edge intensity map and classification network to select the region suitable for fuzzy kernel degradation extraction. After that, ISD (iterative support detection) algorithm is used to iteratively optimize the estimation to get the final fuzzy kernel of each slice, and the result is converted into Energy concerntration (EC) to reflect the clarity of the image. The experimental results show that the method proposed in this paper can realize good region selection effect and effectively improve the ability of extracting image fuzzy kernel, and the method in this paper is ahead of the latest technology.