Edge Computing-Based Real-Time Blind Spot Monitoring System for Tower Cranes in Construction

Most tower cranes require a very complex operating system in order to move objects accurately and safely. However, complex operations may distract the operator, which can lead to harmful accidents. Although existing blind spot monitoring systems have been successfully embedded in cars, simply transferring BSM to construction devices is impractical. In the dynamic construction environment, vehicles and workers work in the same area simultaneously, but the traditional assistant system has a high latency and is unable to provide real-time safety monitoring and alarms. To relieve this problem, this paper designs a YOLO fast-blind spot monitoring system. A YOLO-based system can monitor the tower crane’s blind spot from the bottom of the hook to assist in blind lifting and alert the operator when a potential object is present. This approach relies on edge computing devices to monitor objects’ behavior in an operating blind spot. The results show that this system can detect objects and alert the operator in a potentially dangerous situation with 82.2% precision and an average speed of 110 frames per second (FPS), which fully meet the requirements of a real-time system for dynamic construction environments.