Classification Method for Insulator Temperature Curves Based on Central Axis Extraction by Machine Vision

The infrared detection technology has currently played an increasingly important role in nondestructive online inspection of insulators on overhead power lines. The temperature data in infrared images become essential as the abnormal temperature rise along the central axis of an insulator can be utilized for defect diagnosis. In the first place, a lightweight segmentation method is put forward to extract the central axis of insulator from an infrared image that contains all the necessary thermal information for defect diagnosis, with advantages of reduced storage resources, enhanced storage efficiency and runtime performance when dealing with a large number of infrared images. After that, a classification method for temperature curves by machine vision is proposed, taking advantages of the availability, effectiveness and maturity of the existing image classification model.