Reviewed by Lexie CornerMay 3 2024
A group of scientists from the Universitat Oberta de Catalunya (UOC) have created and evaluated a novel method for identifying asbestos that has not been removed from building roofs despite legal mandates. The study was published as open access in the journal Remote Sensing.
Developed in collaboration with DetectA, the software utilizes artificial intelligence, deep learning, and computer vision techniques on aerial photographs, specifically RGB images, which are widely available and cost-effective.
This represents a significant competitive edge over previous endeavors, which necessitated more intricate and challenging-to-obtain multiband images. The success of this highly scalable project will enable the systematic and efficient monitoring of the removal of this hazardous building material.
Unlike infrared or hyperspectral imaging methods, our decision to train AI with RGB images ensures the methodology is versatile and adaptable. In Europe and many other countries around the world, this type of aerial imaging is freely available in very high resolutions.
Javier Borge Holthoefer, Lead Researcher, Complex Systems Group, Internet Interdisciplinary Institute
Àgata Lapedriza, a researcher with the eHealth Center’s Artificial Intelligence for Human Well-being group (AIWELL) and a member of the UOC’s Faculty of Computer Science, Multimedia, and Telecommunications, is collaborating with Borge Holthoefer on this project. Along with the founders of DetectA, Carles Scotto and César Sánchez, UOC doctoral students Davoud Omarzadeh, Adonis González-Godoy, Cristina Bustos, and Kevin Martín Fernández also contributed to the project.
Using thousands of images from the Cartographic and Geological Institute of Catalonia, the researchers trained the deep learning system to identify which roofs have asbestos and which do not.
A total of 2,244 images were employed, comprising 1,168 positive for asbestos and 1,076 negative. The remaining images were saved for the last test, with the remaining 80 % being used to train and validate the system. Now, the software can identify whether asbestos is present in fresh photos by evaluating various patterns, including the color, texture, and structure of the roofs and the surroundings of the buildings.
The project will benefit rural, coastal, industrial, and urban settings. According to regulations, municipalities were required to conduct surveys of buildings containing asbestos by April 2023, yet not all have completed this task.
Due to their limited availability and high acquisition cost, hyperspectral photographs are not ideal for developing an effective detection method; however, they make asbestos detection easier because they contain many more layers of information. The UOC researchers’ system is the first to use RGB images, which are widely used by cartographic services in many nations and can be captured from aircraft.
“Although these images contain less information, we have achieved comparable results by training the deep learning system well, with a success rate of over 80 %,” explained the CoSIN3 researcher.
Banned for Over Two Decades
Even after asbestos was outlawed for use in buildings more than 20 years ago, it continues to pose a serious threat to public health. More than four million tons of asbestos fiber cement is estimated to be in use in Catalonia alone.
The World Health Organization reports that it results in over 100,000 deaths annually worldwide, primarily from lung cancer but also from pleural tumors and pulmonary fibrosis. The legal deadlines for removing asbestos from public buildings are 2028 and 2032, respectively.
The development of this technological solution will address one of the main challenges in the fight against asbestos: how authorities can determine which roofs contain asbestos so that they can be removed by certified, qualified professionals.
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.
Javier Borge Holthoefer, Lead Researcher, Complex Systems Group, Internet Interdisciplinary Institute
To make the AI system as effective in rural settings as it is in urban and industrial ones, the team is currently investigating ways to increase the training base of the system. This is because the system is slightly more reliable in these settings. After all, it was trained with a larger volume of data from these areas, and also because asbestos wear and conservation differ in rural settings and may be hidden by layers of vegetation.
Journal Reference:
Omarzadeh, D., et al. (2024) Explainable Automatic Detection of Fiber-Cement Roofs in Aerial RGB Images. Remote Sensing. doi.org/10.3390/rs16081342.