A Sensor-Based Method to Detect Near-Miss Struck-By on Construction Site

As a precursor of serious accident, near-miss incidents are of great significance to construction safety management. A real-time detection method for near-miss incidents can improve the performance of construction safety management. However, it is difficult to obtain near-miss data due to the low report rate of near-miss incidents. Aimed at a commonly-seen near-miss incident - struck-by event, based on inertial measurement units (IMU), this research develops a detection algorithm, which includes a decision tree model to monitor the occurrence of near-miss incidents in real-time, and a detection device through the design of a high-wearable hardware. Furthermore, an indoor simulation experiment is designed and conducted. The results show that the proposed method achieves a precision of 97.0%, a recall of 87.8%, a false alarm rate of 3%, and an average response time of 0.43 s. This proves that the method is very effective under the existing data and experimental environment, thus having a potential to support construction safety management in the future.