This project was a UWA CEED project, partnering with Raytheon Australia. The long term aim is to develop a system capable of recognising ships from digital images taken through a submarine periscope, with research to be undertaken over a number of student projects. This was the very first attempt at this problem.
3D Object recognition is not an easy task, however the student developed a number of novel ideas to tackle problems including object recognition, robust feature extraction, and close-to-real-time performance.
A lack of test data led the student to develop an automatic ship generator that can create hundreds of randomly generated 3D ship models in seconds, so that the vessel classification system could be tested. The ship generator will be useful for students undertaking research in this area in future years.
The system was designed to be flexible and extensible, so that future work may build upon, rather than replace, the work done so far.
Results of the work done this year are extremely promising. The student was able to show that the proposed system is capable of fast, accurate classifications from large databases, in the presence of noise and dealing with incomplete data. Research is expected to continue through further student projects in the coming years. |
