London-based Visio Impulse is applying computer vision techniques to help clients and project teams rapidly get insights from site imagery.
Incorporated by Farhad Bazyari in London in June 2017 and launched in February 2018, Visio Impulse applies computer vision techniques to monitor construction and engineering projects. A graduate of Sheffield University, Bazyari completed a PhD in computer science in 2014. He has since been developing applications which can apply artificial intelligence and machine learning to imagery captured via laser-scanning and photogrammetry to assess the accuracy of works and for 3D progress monitoring.
The currently 7-strong company offers services in four main areas:
- Maintenance – identify maintenance issues early and remotely
- Project progress – update project plans and designs automatically as site conditions develop, ensure contractual obligations and timeframes are met, monitor costs and quantities and ensure quality standards and accuracy
- As-built verification – automatically compare as-built drawings and on-site works to ensure accurate hand-over documents, reduce surveying costs and avoid future issues
- Information retrieval – find the right information at the right time for improved project management processes
Petrochemical plant pilot project
The technology was piloted on a major US petrochemical plant that was being constructed by Bechtel. The plant covered an area measuring 3km by 2km and during the two-year project, it was regularly surveyed by weekly aerial drone flights. As well as monitoring progress, Bazyari, right, said the imagery could also be used to validate whether what had been constructed corresponded to what had been designed. Where necessary, the site operator could then authorise any necessary rework and updates to the engineering documentation.
The project also involved some retrospective work, where Visio Impulse was used to check scans of early groundworks and then produce accurate as-built drawings. “These are vital for an operational petrochemical plant,” Bazyari said. “The operator needs full and accurate records, so that it can respond quickly if there is an accident or malfunction at the plant. The complex has an operational service life of 50 years.”
The project highlighted several issues of accuracy in the as-built drawings allowing the client to identify and remedy potentially critical documentation inconsistencies with the contractor. This should avoid potential accidents and interruptions in the future.
Visio Impulse – construction progress validation
While the technology has been proven on a sprawling and complex project, Bazyari says the techniques can just as easily be applied to more compact and conventional schemes such as building construction and infrastructure engineering projects. “Visual validation and verification of work on construction sites is not yet routine in construction,” he says. “But in recent years we have seen growing user of laser scanners and webcams – from ground-level surveys, from drones, and from crane-mounted cameras – to capture progress.”
Once images have been captured and imported, unsupervised Machine Learning automatically generates thousands of labels for faster and improved analysis. The platform’s information retrieval functionality then uses natural language processing to help construction and maintenance teams search for required information among large numbers of documents. By using natural and specific questions, finding the right answer is easier and less time=consuming.
The visualisations created by Visio Impulse can also be used as a ‘time machine’ to help project teams check the sequencing of construction and installation activities. And with BIM increasingly widely deployed, the platform also allows designers to superimpose design models over site captured imagery.
The business is looking to partner with other businesses offering complementary services – for example, drone companies looking to offer advanced computer vision capabilities, and project managers looking to provide new analysis and insight techniques to their owner-operator clients.