Research on Insulator Defect Feature Recognition Technology Based on the Fusion of Multi-Sensor Information

The insulator is a crucial component of the insulation system in power transmission and transformation lines, and it remains widely deployed in operation. As service time increases, insulator defects are more likely to occur due to environmental influences and manufacturing defects, leading to a decline in external insulation performance, which can result in large-scale power outages. Current non-contact detection methods using a single type of information suffer from low efficiency and accuracy, making it difficult to intuitively and effectively assess the condition of insulator defects. Therefore, this paper proposes a method for insulator defect feature recognition based on multi-sensor information fusion. First, the YOLOv11-seg instance segmentation model is improved to achieve real-time segmentation of visible light images of insulators and extract their color features. Ultraviolet pulses are converted into time-frequency spectrograms, and a time-frequency spectrogram processing network is constructed to capture insulator discharge information. Finally, a Neck layer in a convolutional neural network is used to fuse the two types of information, and an insulator contamination state detection model is trained. Experimental results demonstrate that the proposed method meets both accuracy and speed requirements for practical inspections, offering valuable insights for insulator contamination state detection.