Thumbnail Image

Is automatic monitoring accurate enough?

The need for monitoring fish migration has never been greater. With environmental changes and growing aquaculture production in critical areas many rivers are experiencing new and sometimes not-so welcome guests entering the habitats of local species. Some methods of monitoring fish migration such as trapping puts stress on the fish and some non-contact methods such as video surveillance are quite labour intensive. The Riverwatcher is the most widely used fish monitoring system in the world. It was produced to minimize disturbance to the fish and to produce valuable results automatically, such as count, size, direction, PIT-Tag and log several environmental factors such as temperature, oxygen and PH and run the results against migration patterns. It creates automatic reports and sends river information to a data cloud where users can access the real-time information from any web based device. But is automatic monitoring adequate? Only size categorization is fully automatic at the moment. The count is 98% accurate and size measurement is 95% accurate. In rivers with few different species that are in different size and shape this would be adequate. In others where species are similar in size and shape, categorization is a challenge even for the trained eye of experienced scientists. Image recognition software post processing the data can do this to some extent if the image is very good but this is far from accurate at the moment. Until it is, most users would like to identify species and the conditions of the fish themselves by going through the data. The current challenge is to make this process as easy as possible. The future is of course a solution that can identify species and remove unwanted fish in real-time. And that is the Holy Grail our R&D team is working on.