Computer vision applications in construction safety assurance

Weili Fang, Lieyun Ding, Peter E.D. Love, Hanbin Luo, Heng Li, Feniosky Peña-Mora, Botao Zhong, Cheng Zhou

Research output: Journal article publicationReview articleAcademic researchpeer-review

13 Citations (Scopus)

Abstract

Advancements in the development of deep learning and computer vision-based approaches have the potential to provide managers and engineers with the ability to improve the safety performance of their construction operations on-site. In practice, however, the application of deep learning and computer vision has been limited due to an array of technical (e.g., accuracy and reliability) and managerial challenges. These challenges are a product of the dynamic and complex nature of construction and the difficulties associated with acquiring video surveillance data. In this paper, we design and develop a deep learning and computer vision-based framework for safety in construction by integrating an array of digital technologies with multiple aspects of data fusion. Then, we review existing studies that have focused on identifying unsafe behavior and work conditions and develop a computer-vision enabled framework that: (1) considers current progress on computer vision and deep learning for safety; (2) identifies the research challenges that can materialize with using deep learning to identify unsafe behavior and work conditions; and (3) can provide a signpost for future research in the emergent and fertile area of deep-learning within the context of safety.

Original languageEnglish
Article number103013
JournalAutomation in Construction
Volume110
DOIs
Publication statusPublished - Feb 2020

Keywords

  • Computer vision
  • Deep learning
  • Digital technology
  • Safety

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Civil and Structural Engineering
  • Building and Construction

Cite this