Research on personnel detection algorithm for intelligent cross-boundary alarm system in electric power places based on deep learning
Runbo Lu, Ying Zhang, Zeru Zhou, Ke Zeng, Shanqiang Feng, Peixin Xu, Hao Chen, Yungen Liu · IET conference proceedings. · 2025
In recent years, with the continuous development of deep learning, its applications in fields such as image recognition have become increasingly widespread. This study proposes a personnel detection algorithm designed for intelligent cross -boundary warning systems in power facilities. The system, based on deep learning, aims to provide efficient and precise safety monitoring of workers within power sites. The system employs an improved deep learning-based personnel detection algorithm to ensure accurate recognition and detection of workers in two-dimensional images, even in complex operational environments. Improvements to the deep learning algorithm include the introduction of the C3-Res2Block module, which leverages group convolutions to fuse multi-scale features and capture image information at different granularities. Additionally, the C3-Res2Block-DCBAM module is incorporated, utilizing an attention mechanism to assign weighted importance, thereby enhancing key features while suppressing irrelevant and redundant information. Ablation experiments demonstrated that the new modules effectively improved detection accuracy, confirming the positive impact of multi-scale fusion and adaptive attention mechanisms on the personnel detection algorithm, with broad applicability.