Can DIDSON Data Identify Species of Pacific Salmon?

DIDSON data were taken of salmon as they emerged from a fish weir. While passing through the weir, the fish images were captured on video, allowing species recognition. Therefore, the emerging fish could be identified by species. The DIDSON data were analyzed and various features that describe the fish shape and motion were used as input to pattern classification analysis to see if species could be predicted. In particular, fish size and tailbeat frequency looked promising. Unfortunately, many fish did not behave normally after passing through the weir, making analysis difficult. For comparison, video data taken in 1995 of fish that did not react adversely to video equipment were analyzed for tailbeat frequency. Both sockeye and pink salmon were migrating together and species could be identified from the video data. Tailbeat frequency was also measured from these data (76 pink and 77 sockeye) and resulted in frequency distributions that overlapped to a small extent. Pink salmon had a mean tailbeat frequency of 3.1 beats/sec. while sockeye had a mean of 2.3 beats/sec. In addition, Fourier analysis of fish aspect vs. time has been used on DIDSON data to estimate the tailbeat frequency automatically. A comparison of manual vs. automatic analysis of the same fish will be shown.