Abstract: To effectively solve the issues of poor anti-environmental interference and inaccurate defect boundary segmenta-tion in intelligent identification methods of tunnel water leakage.An intelligent segmentation method based on super-resolution reconstruction and triple attention is proposed in this paper,named TR-Unet.In TR-Unet,the Real-ERSGAN super-resolution reconstruction algorithm is used to perform resolution enhancement on water leakage images to strengthen the edge details of the defect regions.Then triple attention mechanism is established by fusing channel attention,spatial attention,and self-attention,and is introduced into the encoder of the Unet network,which enhances the model’s resistance to environmental interference,thus effectively reducing the problems of misdetec-tion and omission caused by the complex environment (light,dark-colored substance interference,appurtenances,etc.) in the process of water leakage identification.Furthermore,water leakage is classified into three categories (green water leakage,dark water leakage,white dry water stained) based on the color and severity of the defect area,and a tunnel water leakage dataset is established based on this.To verify the effectiveness of TR-Unet,five widely used deep semantic segmentation algorithms,Unet,FCN,PSPnet,SegNet,and DeepLabv3+,are selected for comparison.The results show that the mIoU and mPA of the TR-Unet are 83.25% and 87.93%,respectively,which are higher than the other compared models.In addition,TR-Unet has better segmentation effect of water leakage,which is mainly reflected in less false recognition and better edge details.The research results provide a new method for realizing high-precision identification of water leakage in complex tunnel environments.