KelvinSight

BUILDING
THERMOGRAPHY
AT SCALE

Problem to solve

Residential buildings contribute to approximately 20% of greenhouse gas emissions.

An estimated 70% of energy consumption in residential buildings is attributed to cooking, water heating and space heating. Refurbishment of the fabric of residential properties alone has the potential to deliver a 60% reduction in CO2 emissions.

Greenhouse Gas Emissions

Energy Consumption

Potential Reduction for CO2

The solution

By combining machine learning and artificial intelligence with world-class thermal imaging equipment, we will be able to understand and analyse the thermal efficiency of built environments at scale. This will enable data-driven decision making to prioritise funding and interventions for building restoration and upgrades.

Data Gathering

OBJECTIVES

Design, create and curate unique thermal and sensor dataset for build environments.

PROGRESS & ACHIEVEMENTS

  • Designed and developed data acquisition system (camera + sensors) and analysed existing dataset ~17k IR images.
  • Conducted surveys to test multiple system setups (15+ hours / ~1.5m images taken).
  • Adjusted survey routes in collaboration with potential end-users.
  • Established network of potential survey partners / thermography experts.

Data Processing - Machine Learning

OBJECTIVES

Build, train and evaluate machine learning models (computer vision) for classification and object detection.

PROGRESS & ACHIEVEMENTS

  • Designed and deployed data fusion system e.g. to map geo location to images.
  • Developed suitable image pre-processing algorithms to improve data quality and created annotated dataset of ~30k images.
  • Trained and build classifiers for buildings.

Data Reporting - Artificial Intelligence

OBJECTIVES

Evaluate validity and economic viability of proposed solution; gather requirements for system design.

PROGRESS & ACHIEVEMENTS

  • Engaged/interviewed over 50 organisations: housing associations, councils, innovation centres, public services, universities and commercial.
  • Launched and analysed online survey for building and sustainability experts and conducted workshops in collaboration with Construction Scotland IC.
  • Designed reporting dashboards for stakeholder interviews.

Survey COP26 host city to showcase interactive platform and provide data to Glasgow City Council, including most vulnerable areas (fuel-poverty).

  • Increase survey speed through improved camera setup and additional survey modes/partners.
  • Improve machine learning models through improved pipeline, additional data and increase number of classes. Build case-study with selected councils to co-develop AI- driven analytics platforms and integrate with existing datasets, e.g. SIMD.
  • Collaborate with selected partners, e.g. SFRS and NHS, to survey public services real-estate.

WHAT CAN WE HELP YOU WITH?

We are here to help companies who need a long-term software development partner.