Nova Scotia-based Innovasea says it has successfully developed technology to instantaneously detect and count fish using underwater cameras, imaging sonar and artificial intelligence — technology that could help prevent fish kills at hydroelectric dams.
The prototype was field tested this spring at a Nova Scotia Power hydro dam outside Kentville, N.S., where it counted nearly 900,000 fish during an upstream migration of gaspereau — a species of herring.
The aquatic technology company reports the artificial intelligence system was 95 per cent accurate when compared with manual counting done to verify results.
"This is a new capability that gives them a lot more data, a lot more resolution of data and potentially a lot more deep insights as to how fish are moving around the system." said Jean Quirion, Innovasea vice-president of research and development.
An optical camera watched the fish ladder that helps fish get upstream past the 3.4-megawatt White Rock dam. Another camera monitored the dam bypass that allows fish downstream.
Two imaging sonar devices were placed in an outlet canal from the fish ladder to determine if the fish were moving upstream or downstream. Machine learning allowed a computer to count passing fish 24 hours a day with minute-by-minute data on a mobile app.
"It makes insights that you get from the data actionable because they're delivered in real time, all of a sudden you know about them immediately and you have an opportunity to do something about it," Quirion said from Innovasea's Bedford office.
"We all want fish populations to … thrive in our rivers systems while still producing clean hydropower energy. I think technology like this will have a positive impact."
Four years ago, Nova Scotia Power agreed to pay $50,000 into a federal environmental damages fund for causing a large fish kill at the dam in 2017.
Nova Scotia Power said it already takes steps to protect fish passage there. The company currently shuts down the White Rock dam during the spring migration.
"One of the ways the Innovasea study data could be used is to better understand the timing of migration of fish and help us implement these types of mitigative measures," company spokesperson Jacqueline Foster said in an emailed statement.
The company did not provide anyone to interview. Nova Scotia Power did not address a written question asking if the data could enable White Rock to operate safely when it would have otherwise had to shut down.
AI computer will tell the difference between fish species
To further validate the artificial intelligence (AI) technology, an acoustic tracking study to monitor the movement and behaviour of fish tagged around the White Rock dam was undertaken in the spring. The tagging study along with Acadia University is ongoing. The AI deployment will continue during the fall downstream gaspereau run.
Innovasea is currently refining the next generation of its AI technology to identify individual species including endangered species like Atlantic salmon.
"The data being collected will give us a better picture of what species, and how many, are around our facility in real time. We can use that data to further enhance how the facility operates," said Foster.
Why AI works in this case
AI technology that rely on cameras are better suited to narrow spaces like rivers and fish ladders than the ocean where cameras cannot see very far.
The repetitive nature of fish passage is also suited for machine learning, said Jennifer LaPlante of the IBM Deep Sense ocean data hub at Dalhousie University in Halifax. Deep Sense helped design the system.
"It needs to be shown repeated examples of the image over and over for this computer model to actually learn so that it understands what it's meant to look for. And once that happens then it can just do it automatically 24/7 for multiple cameras," she said.
"This is a great example. You have Nova Scotia Power who have a need of monitoring and observing and watching. It's a very repeatable task, but it's very manual and time-consuming and essentially a waste of resources. That's where AI is great. You can go and take a task that a human does and automate it and make it much more efficient and effective and reduce costs."
The fish detection technology developed by Innvoasea makes up about one-third of the $29 million Ocean Aware project funded through the federal Ocean Supercluster innovation fund. The Ocean Supercluster put in $13.74 million with $15.7 million coming from industry partners.
"it's been very exciting for us," says Quirion.
MORE TOP STORIES