|Project Title:||Enabling Technologies based on Computer Vision and Machine Learning for Better Quality Management on Construction Sites|
|Hosting Institution:||The Hong Kong Polytechnic University (PolyU)|
|Abstract:||Quality deviations are common phenomena in the Hong Kong construction industry and they cannot be effectively managed by conventional methods. Quality deviations can cause significant cost overruns and project delays, and even become threats to public health and well-being. The existing practice of multi-tier subcontracting system intensifies the situation as subcontractors may use lower standards in executing tasks. In addition, the discontinuous, dispersed, diverse and distinct (i.e. the four ‘D’) nature of construction sites can easily overwhelm quality inspectors/managers who often need to pay daily visits to multiple sites. Quality inspections are time-consuming and labor-intensive; and manually collected quality records / documentations are often inaccurate and missing out the crucial information. Quality management appears to be post-event remedies. In some cases, inspection procedures were completely ignored, and forms were duly signed without conducting actual inspections.
In the previous ITF project (ITP/036/12LP), the project team demonstrated that real-time location data collected from electronic tags can be transformed into proactive safety warning signals to improve construction safety. In another ITF project (ITP/002/16LP), we developed a cloud BIM-based platform, on which computer vision technologies are adopted to recognize construction progress using site images. With these initial successes, we propose to develop enabling technologies to enhance quality management of construction projects.
Specifically, this project aims to develop computer vision (CV) based technologies with machine learning (ML) to transform the current manual quality management practice to an automatic process. The enabling technologies will analyze site activities; real-time captured by surveillance cameras, and then utilize machine intelligence to determine if there are any quality deviations and/or defects within these site activities.
This study represents a novel development and adaption of CV and ML to liberate human inspectors from tedious site inspections and manual documentation. The developed technology can eliminate the root causes of quality deviations in construction projects. The technology is non-intrusive, and it requires little additional costs (surveillance cameras are already commonly installed in sites) in its implementation. The estimated cost of upgrading an average security surveillance camera to serve quality monitoring is around 300 HK dollars.Therefore, it possesses a great potential to be widely adopted.
Multiple organizations, including Chun Wo Building Construction, Hailong Construction Technology, Harbour Group, and Openplatform Technology, have all expressed strong interests and supports. Two trial projects are provided by the contractors.
|Project Coordinator:||Prof Heng Li|
|Approved Funding Amount:||HK$7.6M|
|Project Period:||01 June 2018 - 31 May 2020|