Artificial intelligence-based recognition and modification of workers' unsafe behavior

The existing research on behaviour-based safety (BBS) research has the following challenges，unsafe behaviour data is not fully utilised，unsafe behaviour analysis ability is insufficient，and unsafe behaviour correction lacks pertinence．Against this contextual backdrop，a framework system for intelligent recognition and correction of workers' unsafe behaviors based on BBS and machine vision was constructed． The proposed framework consists of automated identification of workers' unsafe behaviour，personalised safety training for behaviour modification，and automated prediction and tracking of workers' unsafe behaviour．The framework was applied to Wuhan Lianghu Tunnel Project．The results demonstrate that the proposed framework can identify and predict workers' unsafe behaviour， and therefore enable appropriate training to take place． The framework developed not only informs the health and safety personnel of unsafe behaviours that have occurred on-site but can also be used as an educational tool to inform people about the need to perform their work safely． It can be suggested that our research provides a new direction for exploring BBS management supported by computer vision technologies． © 2022 Huazhong University of Science and Technology. All rights reserved.