Abstract: Cables are one of the three critical components for safe production in lifting machinery, widely used in ports, shipyards, hydropower stations, and other industries. Abnormal cable entanglement accelerates wear and wire breakage, potentially triggering safety incidents that endanger personnel. Currently, cable condition monitoring relies primarily on manual inspections, which are not only inefficient but also carry high risk. Furthermore, existing general-purpose object detection algorithms struggle when processing long-to-width ratio objects like cables. Limited by their field of view and feature extraction capabilities, these algorithms are prone to false negatives or false positives, making it difficult to distinguish subtle anomalies such as back-cable or jump-cable issues. This paper addresses this challenge by proposing an improved YOLOv8n-based cable entanglement detection algorithm aimed at achieving real-time monitoring of cable entanglement states. Leveraging the tubular slender morphology of cables, the algorithm employs a Dynamic Snake Convolution (DSConv) module combined with a PSA attention mechanism. This effectively enhances the model's ability to focus on features of elongated curved objects and localized abnormal regions. Furthermore, integrating the GhostC2f module and HS-FPN module, while replacing the bounding box loss function CIOU with the WIOU loss function, achieves model lightweighting while maintaining accuracy. Experimental results demonstrate that the improved YOLOv8n algorithm exhibits outstanding performance. Compared to the baseline YOLOv8n, the enhanced model achieves an accuracy of 96.7% (a 2.6% improvement) and a mean average precision (mAP50) of 95.2% (a 0.9% increase). Simultaneously, the number of parameters decreased by 57.8%, and the number of floating-point operations decreased by 37.8%. The detection speed reached 185.7. The results demonstrate that this algorithm can accurately identify cable winding states, improving model detection performance while achieving lightweight modeling, enabling deploy0ment in complex environments.