A multi-modal information interaction-based detection method for substation equipment using YOLOv11
Yue Wang, Zhiyuan Pan · Electric Power Systems Research · 2026
Reliable detection and localization of substation equipment under normal operating conditions is paramount for the autonomous inspection of power systems. However, traditional single-modal detection methods often suffer from performance degradation under adverse lighting conditions or complex thermal backgrounds. This paper proposes a robust multi-modal information interaction detection framework based on the state-of-the-art YOLOv11 architecture. To effectively leverage complementary information from visible and infrared modalities, three novel modules are integrated: (1) the Feature Information Extraction and Integration (FIEI) module, designed to capture fine-grained spatial and thermal features; (2) the Multi-modal Feature Shunting and Merging (MFSM) module, which adaptively resolves feature conflicts and synchronizes heterogeneous data; and (3) the Cross-modal Feature Enhancement (CFE) mechanism, which employs attention-based interaction to suppress noise in low-quality images.The experimental results on a self-built multimodal dataset of substations show that the accuracy of the proposed method reaches 91.3 %, which is 15.56 % higher than that of the visible light image detection method and 18.38 % higher than that of the infrared image detection algorithm. Compared with the mainstream image fusion detection methods, the detection accuracy is improved by an average of 10.87 %.While maintaining a relatively low computational complexity, it significantly suppresses the phenomena of missed detection and false detection, showing strong performance for equipment localization and detection in normal operation scenarios.