Construction safety-related research is either management-driven or technology-driven. Many emerging technologies have been investigated by researchers worldwide in different construction management applications. Although there is some limitations, recent advances in deep learning technology has made computer vison (CV) a very active research topic in the field of construction safety management. However, most CV-based studies focus on detecting unsafe behaviour without extending the work to include identifying, locating, and notifying the people involved. In this paper, a framework is proposed for an on-site construction safety management system using Fast R-Convolution Neural Network (CNN)-based computer vison and Bluetooth Low Energy (BLE)-based real-time location system (RTLS). CV can detect onsite entities, understand their spatial relations in a semantic way, and recognize actions of workers and construction equipment. Moreover, the trajectories of moving objects can be tracked, and the next location can be predicted. Combined with the low-cost and easy-to-implement BLE-based RTLS, workers involved in a potential construction hazard will be warned through a mobile application on their smartphones, via loud sounds and vibrations. Experimental studies were conducted at two construction sites to test the CV-based method and the RTLS-based worker identification and warning.