AI real-time detection of cable corrosion in power inspection videos based on a YOLOv10-transformer hybrid network

Abstract

The operation and maintenance of current power infrastructure face critical challenges, such as inefficient cable corrosion detection, strong subjectivity in assessments, and insufficient early-warning capabilities. Traditional manual inspection methods not only demand substantial human resources but are also susceptible to environmental factors and inspector experience, which compromises detection accuracy. Therefore, this study proposes a deep learning model that integrates a multi-scale feature extraction module with an adaptive attention mechanism (MSFEM-AAM). The model employs parallel feature extraction pathways to simultaneously capture global morphological characteristics and localized corrosion details on cable surfaces. A dynamic weight fusion mechanism is incorporated to adaptively integrate multi-scale features, thereby enhancing the model’s discriminative capability for corrosion patterns. Experimental results show that MSFEM-AAM achieves 96.8 % accuracy in corrosion grade classification (Task 2) and 94.7 % mAP in corrosion region detection (Task 1). The single-frame inference time reaches 9 ms with GPU acceleration. These quantitative metrics demonstrate the advanced performance and practical applicability of MSFEM-AAM, contributing to improved automation of corrosion detection and enhanced early-warning capabilities, thus supporting the secure and stable operation of power grids.