Remote Sensing Image Segmentation Algorithm Based on Improved SegNet

High-resolution remote sensing image contains abundant ground object detail information and land cover information, which has great potential for mining and application of high-value data information. Remote sensing image segmentation can provide basic data for the rational planning and utilization of land resources, provide important clues for the exploration and development of mineral resources, and provide decision-making basis for the rational development and utilization of water resources. In order to solve the problems of low data utilization rate and insufficient high-value information mining in the existing remote sensing data processing technology, this paper studies the improvement of SegNet based on ASPP network structure, and focuses on the mathematical model of ASPP and improved ASPP structure. GeLU replaces ReLU to dynamically adjust the nonlinear strength. WHDLD data set was used to compare and analyze the improved SegNet model with baseline models such as “SegNet, FCN, U-Net, DeepLab” through simulation experiments, which verified the superiority of the improved SegNet model and pointed out the shortcomings of the baseline model.