In the September 3 Watermark, we reported on the robotics team at Queensland University of Technology in Australia that developed COTSBot, the autonomous, underwater robot that can hunt, identify and kill Crown of Thorns Starfish (COTS) that are literally eating away at the Great Barrier Reef.
Here we discuss the future of the COTSBot technology with team member Feras Dayoub, a self-described “computer vision and robotics researcher” who is conducting his postdoctoral fellowship with the Australian Centre of Excellence for Robotic Vision at Queensland U.
Watermark: Will you be experimenting with applications of the COTSBot technology to other species or systems?
Dayoub: In this particular project, we are interested only in targeting COTS. But another example of where this detection system can be deployed is in precision agriculture where mechanical weeding or targeted herbicides treatment is preferable. Work using previous vision-based technology has also targeted the detection and counting of Northern Pacific Seastar, a pest that was introduced via ships ballast water. A particular problem was the low-light and turbidity which made image recognition quite difficult.
Watermark: You mention turbidity — this is the natural condition of our local river, the Hudson; there is a lot of “noise” in our system. How well does COTSBot perform under such conditions?
Dayoub: As this system depends on computer vision enabling technologies, visibility conditions are a major factor in determining its performance. In recent, unrelated research we have looked at improving the ability to see through turbid water using real-time computer vision (link here). This could be an option as a pre-processing step in far more challenging visibility conditions. Optical-Acoustic imaging is another answer to such poor visibility conditions. However, this type of sensor requires precise calibration. I can point you here for more information about state-of -the-art technique and imaging sensor: http://www.psmfc.org/tsc2/docs/Project_Profile_Summaries_FINAL.pdf.
Watermark: What about other threats beyond the purely ecological?
Dayoub: We see this COTS system as a nice framework to explore new and more challenging conditions. There are a number of other pests, both current and potential, which are biosecurity risks here in Australia, that we would like to extend the system to, initially as an early warning system.
Watermark: The robotics of COTSBot are impressive, the sensor tech as well. We are interested at Beacon Institute in the future of sensors and real-time monitoring. Give us your insights on what lies ahead.
Dayoub: It is likely the next 5-10 years will see some really exciting new sensors come into use in aquatic systems. These may include olfactory sensors, miniaturisation of mass spectrometers, and even in-situ real-time nutrient sensors. The major challenge would be to get these sensors into a form factor with reliability for installation on small class robotic vehicles. We also believe there is still a lot that real-time computer vision can do to improve monitoring of all sorts of pests, and even change of seabed characteristics.
Visit Feras Dayoub’s website at this link.