A Deep Learning-Based Method for Diagnosing Guano Contamination Status of Transmission Line Insulators

Since the commissioning of high-altitude transmission projects such as the ±400 kV Chaila Line in the Qinghai-Tibet region, insulator flashover faults caused by guano have occurred repeatedly, highlighting the serious pollution flashover risks and insulation safety challenges faced by plateau transmission lines under guano pollution. Timely detection of guano accumulation on insulators is key to preventing such flashovers. To improve detection and segmentation accuracy of insulators and their guano coverage in complex environments and accelerate model inference, this study constructed an insulator guano contamination dataset by capturing images of disassembled plateau insulators and applying image enhancement and synthesis. An intelligent recognition method based on deep learning was proposed to achieve accurate segmentation of guano-contaminated areas and contamination level identification. The model was optimized via lightweight design and attention mechanism addition based on its original architecture, ensuring both fast detection speed and good accuracy. This enables rapid, non-contact assessment of guano accumulation, providing technical support for technicians to analyze insulator status and formulate cleaning and maintenance plans.