Researchers Turn to Computer Vision to Track Down Asbestos Roofs for Safe Removal

Trained on visible-light aerial imagery, the model aims to highlight buildings whose roofs likely contain dangerous asbestos.

Gareth Halfacree
1 month ago β€’ Machine Learning & AI

Researchers from the Universitat Oberta de Catalunya (UOC) and Northeastern University, working with DetectA, have come up with a computer vision system which can be fed aerial photographs and flag roofs which may contain dangerous asbestos requiring safe removal.

"There is currently no protocol or effective system for locating the asbestos that is still out there, because it is expensive and time-consuming to inventorize using people on the ground. Unlike infrared or hyperspectral imaging methods, our decision to train AI with RGB images ensures the methodology is versatile and adaptable," claims Javier Borge Holthoefer, lead researcher on the project. "In Europe and many other countries around the world this type of aerial imaging is freely available in very high resolutions."

Once hailed as a wonder material, asbestos was commonly used as a building material β€” often as an additive in concrete β€” until the discovery that airborne fibers cause cancer, with the World Health Organization estimating asbestosis deaths at over 100,000 globally each year. Many countries have banned its use, but nowhere has eliminated its presence β€” which is where the team's computer vision system comes in.

Using a dataset of 2,244 hand-labeled images of roofs, 1,168 of which were made with asbestos and 1,076 of which were not, from the Cartographic and Geological Institute of Catalonia, the team was able to create a model which can analyze new images taken from aircraft with standard visible-light cameras β€” and not the rarer and more expensive hyperspectral sensors used in prior studies. "Although these images contain less information," Borge Holthoefer admits, "we have achieved comparable results by training the deep learning system well, with a success rate of over 80 percent."

The team's work has been published in the journal Remote Sensing under open-access terms.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles