In this study, we aim to develop an innovative factory safety monitoring system that integrates 3D convolutional neural networks (CNNs) from the field of machine learning with IoT technology to address the safety and efficiency challenges faced by the industrial production sector in the post-pandemic era. By utilizing an improved 3D CNN model, the system can capture and analyze potential safety risks in the factory environment in real time from dynamic video footage, enhancing the understanding and predictive capability of continuous actions and behavior patterns. We demonstrate that 3D CNNs exhibit higher accuracy and dynamic scene analysis capabilities in capturing spatio-temporal features than traditional 2D CNNs. Furthermore, the integration of IoT technology facilitates more efficient data collection, transmission, and real-time analysis, thereby strengthening real-time safety monitoring and decision support. The application and validation of the system in real factory environments have proven its effectiveness in enhancing production line safety and operational efficiency, offering a new solution for the fields of industrial automation and intelligent manufacturing to bolster real-time safety monitoring and decision support.