Cable security state detection based on multi-scale information fusion and distributed fiber optic sensing

Abstract Mechanical operations in proximity to cables present considerable risks to the stability of power transmission and the reliability of energy supply systems. As an advanced intelligent sensing technique, Distributed Fiber Optic Sensing (DFOS) has the capability to detect and identify potential external hazards affecting cables. However, due to the massive volume of sensing data, existing studies primarily rely on classification algorithms to distinguish threat events, lacking efficient end-to-end detection algorithms for differentiation and localization. This constraint results in inefficient information usage and limited real-time responsiveness. To overcome these challenges, this study introduces a high-performance detection algorithm that integrates multi-scale information fusion. First, an attention mechanism called MRFA is proposed to achieve effective feature extraction, which is characterized by its flexibility and multi-receptive fields. Second, an innovative Information Dialysis Module (DM) is proposed to enhance the efficiency of inter-layer information filtering in detection models. Finally, the proposed methods are integrated into an improved YOLOv8 framework. Experimental comparisons across multiple datasets validate the proposed method’s effectiveness and efficiency, demonstrating its capability for real-time surveillance and smart recognition in cable security applications.