In the construction of smart warehousing for power grids, the detection of power fittings on conveyor belts is a critical task. Traditional methods relying on manual labor face limitations such as low detection efficiency, high false detection rates, and high overall costs. This paper proposes a lightweight algorithm based on improved YOLO11n. First, a novel DSCF‐SC module incorporating SCConv enhances the backbone network. This module captures global contextual information—such as spatial relationships between power fittings and conveyor scenes, cross‐scale arrangement patterns, and environmental features—while compressing spatial‐channel redundancies. A parallel micro‐branch preserves fine‐grained details like surface textures, and the dual‐branch fusion improves multi‐scale power fittings recognition. Second, the neck structure is redesigned using Grouped Shuffle Convolution and a slim‐neck paradigm for a lightweight design. Finally, a programmable gradient information (PGI) detection head mitigates information loss in deep networks. Experiments on a self‐built power fittings dataset show the model achieves a mean average precision (mAP) of 74.45%, outperforming the baseline by 1.46%. With a model size of 8.6 MB and an inference speed of 213.37 FPS, the approach effectively improves detection accuracy while ensuring real‐time performance, meeting the requirements for on‐site deployment and supporting automation in intelligent grid warehouses. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.
Zhao et al. (2026) studied this question.