Research on Insulator Defect Detection Based on Computer Vision

Abstract

Accurate detection of insulator defects is essential for ensuring the safe and reliable operation of power transmission lines. However, complex backgrounds, varying lighting conditions, and small defect regions often affect the accuracy of conventional detection methods. This paper presents an improved deep learning-based approach for automatic insulator defect detection. The proposed method incorporates an attention mechanism to enhance feature extraction and reduce interference from complex backgrounds. A multi-scale feature fusion structure is introduced to improve the recognition of defects at different scales, while an optimized bounding box regression loss function is adopted to improve detection accuracy. Experiments are conducted on a public dataset containing 400 insulator images, including broken insulators, self-explosion damage, and flashover defects. The experimental results show that the proposed method achieves a precision of 93.7% and an mAP of 77.8%, representing improvements of 9.4 and 2.3 percentage points, respectively, compared with the baseline model. These results demonstrate the effectiveness of the proposed approach for insulator defect detection in complex outdoor environments.