David Staples: Did the Earth move for you? Irene Cheng’s team can measure it from space

The breathtaking change in computer science can be seen in the shifting focus of University of Alberta computer scientist Irene Cheng over her career.

Cheng started her work in computing science in the early 1990s, before the days of Google or even AltaVista and WebCrawler. Her then-cutting-edge research was to develop a search engine and algorithm that could take a databank of digitized images and find the image that best matched an individual search.

These days Cheng is involved in a new kind of cutting edge, using artificial intelligence and data from satellite radar images to monitor industrial projects and land forms for millimetre-sized shifts in elevation so that both man-made and natural disasters can more easily be predicted and proactive measures taken to avoid human harm and property loss

A Hong Kong native, Cheng joined Hong Kong Bank and Lloyds Bank out of high school, where she worked on bringing in major upgrades to early electronic databanks for these institutions. This led her to study computer science and data bases at the University of Alberta, a world leader in computing science. She came here because her father and brother had previously immigrated to Canada.

After working on her image search algorithm, Cheng worked in private industry for a few years. At that time bandwidth was greatly limited and expensive. Computers were much slower. Cheng’s focus was to find ways to deliver lower resolution images that humans could comprehend, thus making best use of bandwidth.

Any image is composed of pixels, multiple points. A human eye may not be able to see a particular pixel based on how far away they are from the object, either too close or too far. Cheng found the point at which humans could not register the extra detail and set transmission levels to that point. “I looked at what was the minimum I could transmit and still satisfy the user.”

This concept can be applied to many applications, she says, such as 3-D modelling, graphics and animation and gaming. And the issue around bandwidth isn’t going away, Cheng said.

“It’s kind of tug of war because bandwidth is always an issue if the demand keeps increasing. It doesn’t matter how much bandwidth or high speed internet fibre optic there is at the moment — and compared with 10 or 20 or 30 years ago it’s really immense. The demand always outperforms the provision of bandwidth.”

On her latest project, which was put together and partly funded by the federal government, Cheng has teamed up with a Vancouver company, 3VGeomatics.

The company uses radar satellites (there are five such satellites now orbiting the Earth) to emit pulses of radar waves which hit the Earth’s surface and bounce back to the satellite. By measuring the differences in return time after each flyover, the company said it is able to measure ground displacement over time. These radar readings are so accurate that if a rock formation shifts near a major highway or a dam at a tailings pond sinks a few millimetres, or a section of an airport runway is sinking, 3VGeomatics is able to discern it, report it and appropriate action can be taken. “If you know the tendency is to move to a certain direction on a mountain slope then you know very likely a landslide is coming,” Cheng said.

Cheng and her U of A team were called in to help the company move away from humans analyzing the radar data and instead have artificial intelligence programs do the work.

The radar satellite’s record an utterly massive amount of data, but some of that data is contaminated by moisture and dust in the atmosphere. “It disturbs the accuracy,” Cheng said.

One task of the A.I. is to separate out the noise, the corrupted data, from the useful data, the signal that there might have been some movement or shift on the Earth’s surface that could mean trouble.

In the past in order to discern movement, 3VGeomatics needed about 30 satellite flyovers for comparison purposes. By using Cheng’s program and cleaner data, that is now down to just one flyover. “If we trust the data is clean, we can then calculate the elevation change.”

Cheng says she’s greatly enjoyed this work, including the team and collaboration she has on this current project. “I enjoy it very much, especially when working with the students, the current team. They are so dedicated. They work as one and we work very well with the company. Over time we’ve developed trust between us.”


Note from WSOE.Org : This content has been auto-generated from a syndicated feed.