On several of the Columbia River and tributary dams, in order to augment the data from ladder fish counting windows, fish are diverted into off-ladder collection facilities to get additional fish specific data. This, however, usually involves an interruption in upstream migration, temporary holding, the use of anesthetics and manual handling. In this session we will describe an automated system that can be used in these environments to collect more accurate counts, fish measurements and other data at the fish passage structure, while minimizing or eliminating stress and delay normally associated with fish handling.
Whooshh Innovations has developed an advanced fish scanning system, the “FishL Recognition System”, combining advanced imagery, controlled lighting and utilizing recognition algorithms derived from machine learning to produce a data gathering device specifically optimized for “no handling” fisheries management data collection. Unlike typical counting systems that are subject to turbidity and other issues, the system relies on brief “dewatering”, providing much greater image clarity thus enabling automatic, rapid assessments to be made such as size measurements, speciation and adipose fin clip detection as the fish traverses the system.
Imaging/data analysis takes less than a second, and the results can be used in real time to make sorting decisions for separation of species or, for example, hatchery verses wild stocks. In some cases it may be desirable to deny passage to certain species, allowing for removal or return to the tailrace of individual fish. In addition, a permanent record image file is made and logged enabling fish assessment and sorting decision confirmation, validation and additional data mining.
This session will describe the process of developing an automated system like this and the implications for deployment of this kind of technology for active fisheries management.