Abstract: The existence of floating objects (such as plastic greenhouses, plastic mulches and dustproof green nets) along the transmission corridors is one of the main hidden dangers of external damage to transmission lines, and it is important to conduct remote sensing identification of the floating objects for the management the risk of external damage. Most of the existing remote sensing identification studies only focus on one type of floatable objects, and few of them have been carried out in the areas along the transmission corridors. Consequently, a semantic segmentation model based on improved UNet is proposed, which contains Atrous spatial pyramid pooling (ASPP) module, Non-local feature extraction module and Sub-pixel convolution strategy. The area along four transmission corridors in Jiangsu Province is selected as the study area, and the overall accuracy of the experiment is 94.86% with mean intersection over union of 0.6805, which is superior to that obtained by the traditional semantic segmentation model, and the ablation experiments verify the effectiveness of the above three modules. The study shows that it is feasible and effective to utilize the model in this paper to identify floating objects along transmission corridors based on high-resolution satellite remote sensing images, which can provide a decision-making basis for the comprehensive management of the risk of external damage.