Remote sensing identification of easily floating objects along transmission channels based on improved UNet

The floating objects along the transmission channel have a probability of covering the transmission line and causing faults in strong wind weather, which is one of the main external damage hazards of the transmission line. Most of the existing research only focuses on one type of easily floating debris, and there is not much research on remote sensing recognition of green nets. We conducted floating object identification in several transmission channel areas in Jiangsu Province based on high-resolution satellite remote sensing images and deep semantic segmentation models. We propose a semantic segmentation model based on improved UNet. We first annotated 3 types of easily floating debris samples and 1 type of negative samples in the study area through visual interpretation; Then, the proposed model is used for training and testing, and compared with traditional deep semantic segmentation models; Finally, predictions are made throughout the entire study area. The overall accuracy of the experiment is 94.86%, with a mIoU of 0.6805, which is higher than the accuracy of traditional semantic segmentation models. Research has shown that our proposed model is feasible and effective in identifying easily floating objects along the transmission channel, and can provide decision-making basis for comprehensive management of external hazards.