DIN funds sonar image classification R&D for AUVs
The University of Wollongong has led a collaboration with academia, industry and DST Group to develop and investigate the application of convolutional neural networks and advanced machine learning for autonomous underwater mine detection and recognition using sonar systems.
The research is funded by a $124,017 grant from the Sydney-based Defence Innovation Network. The UoW-led team includes Macquarie University, Western Sydney University, DST and Solutions from Silicon Pty Limited.
While deep learning can produce state-of-the-art classification performances in several application domains, it often relies on a large amount of hard to obtain data with which to ‘rain’ the system. This project has increased the understanding of the potential for deep learning to benefit automatic classification of snapshot images containing mine like objects (MLOs), non-mine like objects (NMLOs) or False Alarms as detected by automatic target detection software applied to sonar images. The experimental results indicate the feasibility of the proposed techniques, with a classification accuracy of 98.3%.
These outcomes outline the potential to incorporate these approaches into automatic target detection software that has been commercialized by DST Group in collaboration with the Solutions from Silicon.
Further information from the Defence Innovation Network here