Machine learning helps the anti-shark drones figure out what’s a shark and what’s not.
Last year, there were 26 unprovoked shark attacks on humans in Australia, two of them fatal. Researchers in Australia are working towards reducing those numbers with a drone capable of detecting sharks through a machine learning algorithm.
Building anti-shark drones
The anti-shark drones are the result of partnership between The Little Ripper Group, a private company focused on search and rescue drones, working with researchers from the University of Technology Sydney’s (UTS) School of Software. Little Ripper provides the drones, UTS provides the algorithm.
“The automated system for detection and identification of sharks in particular, and marine life/objects more generally, was developed using cutting edge deep neural networks and image processing techniques,” says Professor Michael Blumenstein, Head of the School of Software in the Faculty of Engineering and IT in a statement.
That detection system was built using image processing techniques that examined live video feeds in real time to detect sharks. Using an object detection system called a Region based-Convolution Neural Network (RCNN), UTS is able to detect sharks with 90 per cent accuracy. It filters them through myriad images of other marine life and human surfers and swimmers.
What’s in a drone
Little Ripper’s drones, a fleet made up of US and German-built UAVs, are geared towards sharp visuals. The Vapour 55’s battery can last up to an hour on a single charge. The drone is fitted with a gyro-stabilised digital camera with FLIR thermal imaging and an optical telephoto zoom.
“This system will help make beach recreation much safer and is a major milestone in addressing shark attacks with very real ability to save a life,” says Eddie Bennet, Little Ripper Lifesaver‘s CEO. New South Wales and Queensland surfers can expect the drones to start their detection process in September.