SharkSpotter drones deployed this season
Summer skies in 2017/18 will see more Little Ripper drones using the UTS-developed SharkSpotter software to warn of possible shark attack.
A fleet of drones featured at the recent Raising of the Flags to mark the official launch of the season by Surf Life Saving NSW, so will be more visible on city and regional beaches as the weather attracts more people into the water.
The drone is now being fully deployed with the shark detection algorithm developed by Dr Nabin Sharma from the School of Software, a member of the UTS team led by Professor Michael Blumenstein. The algorithm can distinguish between common marine life such as whales, dolphins, rays and seals as well as sharks that menace swimmers and surfers. Nabin is currently ‘fine-tuning’ the algorithm and the next version will extend the 16 current categories to include turtles.
“The most recent tests at Port Macquarie during Spring mid-term break have given us more information and understanding of the optimum conditions for the drone to effectively capture and relay aerial images from the sea,” he said.
“We are trying to reduce motion blur to deliver more accurate detection results back to shore.”
At Port Macquarie, there were multiple flights at 30-minute intervals which allowed him to collate data on flight heights and other conditions which can improve the zooming capabilities of the on-board high-resolution camera. During the visit, one shark was spotted close to shore in the early am, and a lot of whale activity as the seasonal migration south to the Antarctic is underway.
“SharkSpotter is an application which supports the general premise that artificial intelligence is meant to assist humans. 90 per cent visual accuracy in shark detection would be very high – and human ability is much less. So if we can get the drone to the best level of accuracy then it will certainly help humans to improve detection rates, and create safer conditions for water users,” he said.
The real-time visual information relayed to shore is interpreted by humans, who have the ultimate decision-making role.
“Information is sent to a control station on the beach where a human responder will have final say on what actions to take – i.e. monitor the shark to see if it moves away from swimmers or, if it appears to become a direct threat, to sound alarms and advise those in the water that they need to evacuate,” he said.
The algorithm used thousands of images from the captured video, and the Department of Primary Industries has supplied more stock which, when annotated, will further improve the system’s capabilities.
Local councils and surf lifesaving clubs have shown huge interest in having a Little Ripper to patrol beaches in their jurisdictions. An interview with Nabin by Reuters Television took the story and the product to a global market – media coverage in landlocked countries in Europe and Africa are a testament to the fear and fascination of sharks, even where no threat is present.