Bringing innovation to industry: Using AI for predicting disasters

    UAlberta computing scientists are applying the power of machine learning to solve problems in industry.

    By Katie Willis on November 5, 2018


    A machine learning model developed by University of Alberta scientists is improving efficiency and accuracy for one wide-area monitoring company that uses data to predict natural and man-made disasters around the world.

    The computing science team, led by Irene Cheng, adjunct professor in the Department of Computing Science, is working in collaboration with 3vGeomatics (3vG) to clean data and automate quality assessment. 3vG, a Vancouver-based company, uses satellite images to study Earth’s surface and detect any changes that could indicate the onset of natural or man-made disasters, like landslides and earthquakes. Many activities, like oil exploration and extraction and mining, can affect the ground surface.

    “The need for this data is growing, and the data is becoming more robust,” explained Cheng, director of the multimedia masters program.

    “3vG saw the need to optimize their systems by minimizing the need for human intervention and increasing the speed for data processing. This means creating software that can automatically tell the quality of images, without human intervention, with the goal of speeding up the process.”

    Enter UAlberta computing scientists, who have worked with 3vG for the last 18 months to develop a machine learning model to determine data clarity, a significant challenge with satellite images. The research team included three PhD students, Navaneeth Kamballur Kottayil, Subhayan Mukherjee, and Alvin Sun.

    “The most important thing is understanding the data that we’re working on,” said Sun. “We needed to transform the data to make the model able to solve the problem. We needed to integrate machine learning to the traditional techniques already being used by 3vG.”

    “In a sense, what we are doing is trying to incorporate their knowledge into the framework that we built, rather than making it generic,” added Mukherjee. “Scalability is the main thing we want to accomplish here for 3vG.”


    The Wide Area Monitoring System (WAMS) Project is funded by Consortium of Aerospace Research in Canada (CARIC) and Mitacs, the first project of its kind of be funded in Western Canada. With the first completed project under their belt, the UAlberta research team plans to continue work with 3vG, with support from CARIC, over the next three to five years to scale up the findings to other challenges facing the industry.