The supervision of electric power safety production is typically dominated by manual inspection approaches

which makes it difficult to meet the requirements of accuracy

objectivity

and real-time responsiveness. With the introduction of computer vision technology

intelligent recognition approaches have gained widespread attention. However

existing electric power personnel behavior recognition algorithms

which rely on single-frame images

inevitably suffer from a high rate of false and missed detections due to the insufficient consideration of motion continuity and contextual information. Therefore

this paper proposes an electric power personnel behavior recognition algorithm based on multi-object sequence capture and spatiotemporal feature extraction. The algorithm obtains personnel behavior sequences through a behavior sequence capture module and extracts spatiotemporal information of behaviors by using a spatiotemporal feature extraction module. It has been validated on public and electric power datasets

achieving an overall accuracy of over 88.6% for typical categories of electric power personnel behaviors. Meanwhile

the algorithm remains amenable to further improvements and can serve as a foundational framework for future related research.