News + Publications

Automating repetitive tasks using machine vision

To make long-term predictions about full-scale deposits, Coanda and Suncor Energy Inc. developed a geotechnical centrifuge. Settling is governed by gravity; spinning the samples simulates an increased gravity and, thus, increases the settling rate. Photographs collected during these centrifuge runs are analyzed to determine the sample’s settling behaviour.

A critical requirement in understanding settling is accurately tracking the solids-water (SW) and air-water (AW) interfaces throughout a sample run. Manual identification of these key interfaces in thousands of images is a monotonous and time-consuming task. Using machine vision and image analysis techniques, Coanda has automated this process to obtain results in under a minute.

The interfaces are identified using a custom edge detection algorithm and are tracked through every image captured during a centrifuge run.

In this video of a 10 cm treated analog sample, the scaled SW and AW interface positions of a 5 m deposit are plotted with respect to the real-time evolution.