AI-based Video Analytics for Progress Monitoring

Traditionally, progress monitoring of construction projects is often influenced by human factors, as project steps – inspection, recording and interpretation – are done manually by different stakeholders. The resulting reports are often not reliable and require further scrutiny by supervisors and higher-level managers.  

Reliable real-time progress monitoring is important to identify issues and potential delays in a timely manner, so that responses like manpower redeployment can be planned and executed to resolve them.

Surveillance tools such as CCTV are being adopted more widely on the construction sites, but these tools have limited usefulness due to the reliance of humans to carry out the remote monitoring.

We are interested in AI-based video analytics solutions that can use the footage from existing or new CCTVs to produce useful insights on on-site productivity. Instead of having site personnels to perform frequent checks on the site, the solution would intelligently monitor the site and alert users of any issues or anomalies.     

Stage completion and floor cycle are the priority areas for progress and productivity tracking.  The tracking of metrics related to manpower, machinery, logistics and material handling are also of interest. 

The solution should fulfil the following specifications in order to be considered successful:

• It must integrate with the existing CCTV system. The majority of Tiong Seng’s current projects utilise video surveillance systems from Dahua Technology.
• It must be able to accurately identify and track the following Reinforced Concrete work activities to quantify the stage completion of a floor as the percentage of completion: 1) Reinforcements works, 2)Formwork installation, 3) Concreting works and 4) Formworks removal
• It must be able to compare the actual progress against projected progress based on a seven or ten-day floor cycle time, and identify issues and potential delays. 
• It must include methods to boost connectivity to allow for real-time monitoring and analytics.  
• The accuracy of the models for the various use cases should be at least 85%.
• It should be able to analyse the duration taken for different stages completed within a floor.

The following are some additional requirements that could further enhance the proposed solution

• The solution could include APIs for interoperability with other enterprise tools and platforms.
• It could be compatible with Building Information Modelling (BIM) to allow for BIM models to be overlaid on real-time CCTV footage or captured images.
• It could involve the installation of new surveillance tools if it can boost the accuracy of the models in a cost-effective manner.  

‍

The solution must include a digital platform that analyses the data captured, visualises the data on a dashboard, produces daily progress reports, and alerts key stakeholders on issues and delays. The platform should integrate the existing live CCTV streams through a RTSP setup.

‍

An AI-based video analytics solution can monitor and track structural work as well as formwork progress from the plan view and elevation view of a construction site. The solution allows for early identification of delays and risks that could cause budget overruns.

After selection 

• Historical CCTV feeds and site progress photos
• Installation of new CCTVs at testbedding sites.