Thermographic survey automation

About the project

External thermographic surveys are non-intrusive, quick and reliable for detecting and understanding defects that cause poor building fabric performance. Qualitatively assessing the condition of the building elements is useful for understanding the potential causes of energy inefficiency, and thus what solutions are available to improve it. However, the information processing tasks which take place after the data has been captured can take competent thermographers as much as 5 times as long as data capture. This limits the availability and accessibility of thermography for assessing building energy efficiency on a large scale. Maesin worked with partners to develop and test software that automates thermographic surveys based on thermal images of building exteriors, using industry-standard data capture processes.

What has the project successfully delivered?

The team have developed a fully functional web application which automates and augments the creation of survey reports from raw thermal images. The software makes it easy to create accurate digitised reports for a portfolio of properties in a highly scalable way, enhancing productivity, reliability and accessibility. The tool is ready to be used in commercial operations starting this winter heating season.

What did the project achieve?

• Worked with qualified thermographers to define requirements for a workflow automation tool for creating reports from raw thermal data.
• Curated a training dataset based on past external thermographic surveys of residential properties.
• Followed thermographers on residential surveys in order to test assumptions and discover unknown unknowns about the data generation process.
• Developed AI models which predict condition ratings of building elements from external thermal surveys.
• Developed a web application implementing the AI models into a workflow automation tool for thermographic survey reports.
• Worked with qualified thermographers who tested the tool and provided feedback, helping to iteratively improve the tool.

Key lessons learned

The ability to verify and override the AI’s recommendations was a critical feature in order for professional thermographers to adopt the tool. This meant that the web application interface had to be designed in a way such that the thermographer still has all the information they need in order to verify or amend the outputs. The team also discovered that for certain types of content, in order for automation to be feasible, there might need to be changes in the way the data is captured. Testing is planned for the proposed changes in data capture techniques this winter heating season.

Next steps

Maesin is looking to work with surveyor partners for stock condition surveys using thermal imaging. They are interested in engaging with the end users of stock condition surveys, such as housing associations and local authorities, to establish how the datasets our tool can unlock could support retrofit programme planning and delivery. Maesin are interested in collaborations with other companies that provide or use building datasets.

Project partners

• iRed Limited