Fish-counting A.I. developed in N.S. may assist avert fish kills at hydro dams

Nova Scotia-based Innovasea says it has efficiently developed know-how to instantaneously detect and rely fish utilizing underwater cameras, imaging sonar and synthetic intelligence — know-how that might assist stop fish kills at hydroelectric dams.

The prototype was discipline examined this spring at a Nova Scotia Energy hydro dam exterior Kentville, N.S., the place it counted practically 900,000 fish throughout an upstream migration of gaspereau — a species of herring.

The aquatic know-how firm experiences the substitute intelligence system was 95 per cent correct in comparison with handbook counting finished to confirm outcomes.

“This can be a new functionality that provides them much more knowledge, much more decision of knowledge and probably a lot extra deep insights as to how fish are shifting across the system.” stated Jean Quirion, Innovasea vice-president of analysis and growth.

An optical digital camera watched the fish ladder that helps fish get upstream previous the three.4-megawatt White Rock dam. One other digital camera monitored the dam bypass that permits fish downstream.

Two imaging sonar gadgets had been positioned in an outlet canal from the fish ladder to find out if the fish had been shifting upstream or downstream. Machine studying allowed a pc to rely passing fish 24 hours a day with minute-by-minute knowledge on a cell app.

Prompt outcomes

“It makes insights that you just get from the information actionable as a result of they’re delivered in actual time, swiftly you already know about them instantly and you’ve got a chance to do one thing about it,” Quirion stated from Innovasea’s Bedford workplace.

“All of us need fish populations to … thrive in our rivers methods whereas nonetheless producing clear hydropower vitality. I believe know-how like this may have a optimistic influence.”

4 years in the past, Nova Scotia Energy agreed to pay $50,000 right into a federal environmental damages fund for inflicting a big fish kill on the dam in 2017.

Nova Scotia Energy stated it already takes steps to guard fish passage there. The corporate at the moment shuts down the White Rock dam throughout the spring migration.

“One of many methods the Innovasea examine knowledge might be used is to raised perceive the timing of migration of fish and assist us implement all these mitigative measures,” firm spokesperson Jacqueline Foster stated in an emailed assertion.

The corporate didn’t present anybody to interview. Nova Scotia Energy didn’t tackle a written query asking if the information may allow White Rock to function safely when it could have in any other case needed to shut down.

AI laptop will inform the distinction between fish species

To additional validate the substitute intelligence (AI) know-how, an acoustic monitoring examine to observe the motion and behavior of fish tagged across the White Rock dam was undertaken within the spring. The tagging examine together with Acadia College is ongoing. The AI deployment will proceed throughout the fall downstream gaspereau run.

Innovasea is at the moment refining the subsequent era of its AI know-how to establish particular person species together with endangered species like Atlantic salmon.

“The information being collected will give us a greater image of what species, and what number of, are round our facility in actual time. We are able to use that knowledge to additional improve how the ability operates,” stated Foster.

Why AI works on this case

AI know-how that depend on cameras are higher suited to slim areas like rivers and fish ladders than the ocean the place cameras can’t see very far.

The repetitive nature of fish passage can also be suited to machine studying, stated Jennifer LaPlante of the IBM Deep Sense ocean knowledge hub at Dalhousie College in Halifax. Deep Sense helped design the system.

“It must be proven repeated examples of the picture time and again for this laptop mannequin to really be taught in order that it understands what it is meant to search for. And as soon as that occurs then it could simply do it routinely 24/7 for a number of cameras,” she stated.

“This can be a nice instance. You’ve gotten Nova Scotia Energy who’ve a necessity of monitoring and observing and watching. It is a very repeatable activity, nevertheless it’s very handbook and time-consuming and basically a waste of sources. That is the place AI is nice. You’ll be able to go and take a activity {that a} human does and automate it and make it way more environment friendly and efficient and cut back prices.”

The fish detection know-how developed by Innvoasea makes up about one-third of the $29 million Ocean Conscious mission funded via the federal Ocean Supercluster innovation fund. The Ocean Supercluster put in $13.74 million with $15.7 million coming from trade companions.

“it has been very thrilling for us,” says Quirion.