AI System Revolutionizes Asbestos Detection from Aerial Images
A game-changer AI-based system by the creative researchers from the Universitat Oberta de Catalunya (UOC) will change how asbestos is detected on building roofs and how the monitoring process is carried out. Through this innovative approach based on deep learning and computer vision technologies, which analyze the RGB remote sensing photographs that are commonly accessible, a solution is provided that is inexpensive and scalable, allowing the addressing of a critical problem like this.
AI overcoming conventional limitations
The old asbestos practices required slow and expensive hyperspectral imaging, which made large-scale facilitation difficult. The team at UOC uses high-resolution RGB images, which are readily available due to the increasing use of aerial mapping services everywhere. The system under development focuses on re-used VNIR and NIR imagery as supplementary aiding information.
“Our approach is really efficient because of the specialty and cost-effectiveness of producing the AI with RGB images.“ Javier Borge Holthoefer, the main researcher at UOC’s Complex Systems group (CoSIN3), explains this. He adds that this high-resolution aerial imaging is free in the U.S. and most other countries.
The researchers started building the system based on material from the Cartographic and Geological Institute of Catalonia, which consisted of thousands of aerial photographs. The deep learning model was systematically educated to pick up the patterns, colors, textures, and structural attributes of the intensive roof ramming patterns in asbestos-containing roofs from the roofs that didn’t.
Multifaceted applications
Such a team utilized a thorough validation framework. In fact, the image dataset was formed of 20% for the primary test. The outcomes were incredibly positive, with the AI system reaching a precision level of more than 80% for detecting asbestos on the roof.
Systems’ versatility is not exclusive to big urban, industrial and coastal areas, rather they can be used everywhere – in the city, in the industry, by the secession and on the field. Authorities can use it for comprehensive surveys appending the safe removal of the asbestos-containing materials from both public and private buildings.
By now, the use of asbestos for construction is regulated and even banned in a number of countries, and still, it represents a tremendous risk internationally. The World Health Organization has registered more than 100,000 death cases that are associated with several lung cancer types, pleural tumors, and pulmonary fibrosis.
It is shocking to see that in Catalonia, the amount of asbestos still inside the buildings is estimated to be more than 4,000,000 tons of asbestos fiber cement. The authorities set a deadline of 2028 for the removal of asbestos in public buildings and 2032 for private buildings as the target, meaning that this issue needs to be resolved with immediate action.
Although the AI system does showcase advanced capabilities in urban and industrial areas, the experts have identified the need to broaden the training base by ensuring it includes more diverse environments. Despite the urban-centric tendencies of existing models, they proffer a broader variety of data for both rural and peri-urban environments.
Read More
AI System Revolutionizes Asbestos Detection from Aerial Images
A game-changer AI-based system by the creative researchers from the Universitat Oberta de Catalunya (UOC) will change how asbestos is detected on building roofs and how the monitoring process is carried out. Through this innovative approach based on deep learning and computer vision technologies, which analyze the RGB remote sensing photographs that are commonly accessible, a solution is provided that is inexpensive and scalable, allowing the addressing of a critical problem like this.
AI overcoming conventional limitations
The old asbestos practices required slow and expensive hyperspectral imaging, which made large-scale facilitation difficult. The team at UOC uses high-resolution RGB images, which are readily available due to the increasing use of aerial mapping services everywhere. The system under development focuses on re-used VNIR and NIR imagery as supplementary aiding information.
“Our approach is really efficient because of the specialty and cost-effectiveness of producing the AI with RGB images.“ Javier Borge Holthoefer, the main researcher at UOC’s Complex Systems group (CoSIN3), explains this. He adds that this high-resolution aerial imaging is free in the U.S. and most other countries.
The researchers started building the system based on material from the Cartographic and Geological Institute of Catalonia, which consisted of thousands of aerial photographs. The deep learning model was systematically educated to pick up the patterns, colors, textures, and structural attributes of the intensive roof ramming patterns in asbestos-containing roofs from the roofs that didn’t.
Multifaceted applications
Such a team utilized a thorough validation framework. In fact, the image dataset was formed of 20% for the primary test. The outcomes were incredibly positive, with the AI system reaching a precision level of more than 80% for detecting asbestos on the roof.
Systems’ versatility is not exclusive to big urban, industrial and coastal areas, rather they can be used everywhere – in the city, in the industry, by the secession and on the field. Authorities can use it for comprehensive surveys appending the safe removal of the asbestos-containing materials from both public and private buildings.
By now, the use of asbestos for construction is regulated and even banned in a number of countries, and still, it represents a tremendous risk internationally. The World Health Organization has registered more than 100,000 death cases that are associated with several lung cancer types, pleural tumors, and pulmonary fibrosis.
It is shocking to see that in Catalonia, the amount of asbestos still inside the buildings is estimated to be more than 4,000,000 tons of asbestos fiber cement. The authorities set a deadline of 2028 for the removal of asbestos in public buildings and 2032 for private buildings as the target, meaning that this issue needs to be resolved with immediate action.
Although the AI system does showcase advanced capabilities in urban and industrial areas, the experts have identified the need to broaden the training base by ensuring it includes more diverse environments. Despite the urban-centric tendencies of existing models, they proffer a broader variety of data for both rural and peri-urban environments.
Read More