Computer vision for behaviour-based safety in construction: A review and future directions

Weili Fang, Peter E.D. Love, Hanbin Luo⁎, Lieyun Ding

Abstract

The process of identifying and bringing to the   fore people’s unsafe behaviour is a core function of implementing a   behaviour-based safety (BBS) program in construction. This can be a   labour-intensive and challenging process but is needed to enable people to   reflect and learn about how their unsafe actions can jeopardise not only   their safety but that of their co-workers. With advances being made in   computer vision, the capability exists to au- tomatically capture and   identify unsafe behaviour and hazards in real-time from two-dimensional (2D)   digital images/videos. The corollary developments in computer vision have   stimulated a wealth of research in construc- tion to examine its potential   application to practice. Hindering the application of computer vision in   construction has been its inability to accurately, and generalise the   detection of objects. To address this shortcoming, develop- ments in deep   learning have provided computer vision with the ability to improve the   accuracy, reliability and ability to generalise object detection and   therefore its usage in construction. In this paper we review the devel-   opments of computer vision studies that have been used to identify unsafe   behaviour from 2D images that arises on construction sites. Then, in light of   advances made with deep learning, we examine and discuss its integration with   computer vision to support BBS. We also suggest that future computer-vision   research should aim to support BBS by being able to: (1) observe and record   unsafe behaviour; (2) understand why people act unsafe behaviour; (3) learn   from unsafe behaviour; and (4) predict unsafe behaviour.

Key words: Unsafe behavior; Computer vision; Deep learning; Convolutional neural network