FishView Final report public summary - Computer Vision...

FISHVIEW

Assessing fish passibility using a robotic fish sensor and

hydrodynamic imaging

Final report Publishable summary

Prof. Maarja Kruusmaa, Tallinn University of Technology Tel: +372 5183074 Email: Maarja.kruusmaa@ttu.ee

Project website address: http://www.ttu.ee/faculty-of-information-technology/centre-for-biorobotics-2/science-9/projects-14/mojor-projects/fishview/fishview-2/

The life-cycle of migrating fish such as trout and salmon depends onenvironmentalconditionsintheBalticSeaanditsconnectingrivers.Upstream migration brings fish into rivers, where conditions are suitable asspawninggrounds,andcontainthenecessaryhabitatsforjuvenilefish.Thusthelife-cycleofmigratoryspeciesheavilydependsonthepassiblityfromtheseatorivers. Establishing and maintain favorable environmental conditions formigrating fish has proven internationally to be a challenging task. Particularlycomplex is theproblemofpassingbarriers suchasdamsandweirswhere thefishwaystructuresprovideawaytopasstheobstructionsandreachthespawninggrounds.However,themajorityoffishpassesarenotfullyfunctional.Therefore, the main goal of the FishView projectwork is to provide a robustmethodologytoimprovefishpassibilityintheriverbasinsoftheBalticSea.FISHVIEWusesanovelapproachtogiveresearchersthe“insideview”ofariverpass.Wedesignadevicethatexperienceswaterflowssimilartothelaterallinesensingorganoffish.Thisdeviceisimmersedinriverfishpassesandrecordsthesignalsintheflow.Thesignalsareanalyzedwithmachinelearningmethods.Astheresult,wehopetobeabletotellhowtobuildfishpassesthatallmigratingfishcanuse.ThemainresultsofFISHVIEWprojectare:

Development of the flow sensitive probe inspired by fish lateral linesensingcapableofsensingpressurewithadistributedarrayofpressuresensorsat high frequencies. This allows recording small pressure fluctuationssimultaneously around a fish like body and perceive the flow from “fishperspective”.

Signalprocessingmethodsfromthosesensorprobesforestimatingflowproperties,suchasvelocityanddirectionbutalsoturbulencerelatedproperties(such as for example turbulence kinetic energy) as well as more abstract“signatures”oflocationsinflowforclassification.

Computersimulationsoftheflowinfishpasses,thatuseoursensordataandsimulatetheflowpropertiesofacomputermodelofafishpass.

We have used our device formeasuring flows in 3 different fishpasses,analysedtheflowconditionsandsuggestedimprovements.