NT Fisheries have been monitoring the Greater Darwin Region reef fish protection areas for the last four years – collecting over 400 hours of video using baited remote underwater video systems (BRUVS). Processing this video, frame by frame to identify fish species, is time and labour intensive.
In collaboration with Microsoft, NT Fisheries has researched the application of computer vision and machine learning to automate the process of fish identification. A successful prototype was built to locate fish within a video frame with greater than 90% accuracy and identify key species with greater than 75% accuracy.
This is a demonstration of remarkable technology given some of these fish images appear for only a fleeting period and often in highly turbid water.
Broader applications of the technology are also being investigated for monitoring bycatch on commercial vessels and for biosecurity monitoring purposes.