An Automated Fish Migration Pattern Monitoring System Using Shape Descriptors for Pattern Recognition

Fish ecology and environmental research scientists have an urgent need for an automated fish identification system for biological research. Many fish species travel great distances during their life cycle. As they travel past a narrow passage on the rivers or dams that are equipped with a fish ladder, their species, size, total number, and the time can be recorded for migration pattern and population studies. These are very important data for biologists, especially for the research of endangered species and effects of environmental change. Manual monitoring and recognition of fish 24 hours a day or reviewing hours of video tapes recorded at fish passageways during the entire migration season is a very tedious task and, frequently results in unreliable data. At places where monitoring facilities are not available, fish must be captured, measured, and counted manually. This invasive method, although not preferable, may be the only option available. Tests conducted in Phase I show very promising results and prove that such an automated system is feasible. This Phase II project proposes to improve the fish tracking and recognition algorithms, conduct extensive field testing in varying conditions, construct an underwater camera system, and develop and prepare to market a commercial product.