A Leak Detection Method for Heat Network Pipes Based on YOLOv5 and Automatic Region Growing Algorithm

With the development of low-carbon energy in industry, heat network pipes has become widely popular as a key piece of equipment for transporting steam energy. However, due to prolonged exposure to the external environment, pipes are prone to damage and rupture and require prompt leak detection and repair. The current leak detection methods still relies on manual inspection, but it has several issues, including incomplete coverage, low accuracy, and a lack of timely detection. As of yet, there is no perfect leak detection methods to meet the demand. To efficiently detect leak faults in heat network pipelines and solve the above problems, this paper proposes a real-time detection method for heat network pipelines leakage based on UAV infrared images is proposed. The method integrates the YOLOv5 model for heat network pipelines identification and introduces a two-stage automatic region growing algorithm to locate the pipe leakage and segment the affected area. Experimental testing confirms the effectiveness of this method in achieving accurate detection. Finally, we developed an application for the inspection system using PyQt, which integrates the YOLOv5 model and leak detection model into the system, achieving the functions of interaction and exporting inspection reports. This inspection system can provide technical support for the inspection of heat network pipelines.