Heavy rainfall and torrential downpours caused by climate change can lead to severe flooding issues near construction sites, resulting in human casualties and property damage. This study developed and implemented an AI-based rainfall safety system using CCTV footage to prevent safety accidents related to flooding near construction sites. The system employs advanced AI object detection technology and image segmentation techniques to detect flooding in real-time and monitor changes in rainfall and water puddle levels to predict risks. The study found that the system accurately detects the formation and level changes of water puddles, contributing effectively to accident prevention through timely warnings. The system developed in this study enhances the safety of areas near construction sites and has the potential for application in various industrial sites. It is expected to become an essential component of smart construction technology, significantly improving the efficiency of accident prevention and safety management.