Real-Time Analysis of Dangerous Actions in Video Surveillance Systems Based on Cloud Computing
Mingzhao Lu, Yonggang Xu · 2025
With the increasing safety management demands in high-risk environments such as industrial and construction sites, real-time identification and warning of dangerous behaviors have become a significant research topic in video surveillance. This paper designs a real-time analysis system for hazardous actions based on cloud computing, utilizing a three-layer architecture comprising edge, network, and cloud components. By combining an improved 3D convolutional neural network and a multi-scale feature extraction strategy, the system achieves lightweight target detection and hazardous action recognition. The system employs a microservice architecture, supporting concurrent processing of multiple video streams and real-time alarm notifications. Experimental results indicate that this system significantly outperforms traditional systems in terms of processing performance, resource utilization, and scalability, effectively meeting the safety monitoring needs of high-risk environments.