Deep learning could help prevent salmon escapes in choppy seas

Searching for damage or holes in marine cage nets is one of the most important routine operations in the aquaculture industry. Major breakthroughs in both machine learning and deep learning have provided us with automated methods for network auditing. Credits: SFI Exposed/SINTEF

Artificial intelligence can be of great benefit underwater, and SINTEF, in collaboration with research center SFI Exposed, is developing systems to help improve fish farm safety and security in harsh marine conditions.

You may have heard of deep learning already. It is an important process in the field of machine learning that allows computers to expand their capabilities in response to the information provided to them. By exposing a computer brain to concepts it needs to understand, it can be taught to recognize contextual information. In a data structure modeled on the human brain, devices are connected to form what we call a neural network. Such networks consist of many layers that interact to enable what we call deep learning.

We have used this concept to develop and improve systems that can detect irregularities in salmon cage nets in fish farms. We do this by feeding a digital neural network with images that show the building blocks of an intact network. These images then allow the program to react when it observes something unusual, such as a hole.

Escaping fish is a major problem

Frequent inspection of marine cage nets suspended under water is intended to help prevent fish from escaping. Escaped farmed salmon can carry disease and can also be disturbed and involved in the spawning process by migrating upstream with wild fish along rivers. These factors only serve to weaken wild salmon populations, so it is in everyone’s interest to prevent escapes from marine cages.

The Norwegian government recently reported that it plans to introduce stricter technical requirements for the escape-proof design and operation of fish farms. These new regulations make us doubt that we need to think innovatively and may offer an opportunity to implement or further develop some of the concepts we are currently working on at the SFI Exposed research center.

Current standard operations include the use of a camera-mounted, remotely operated vehicle (ROV) that is manipulated by an operator who examines images that are launched into a cage and then transmitted back. It is difficult for an operator to concentrate for hours on monotonous video footage of underwater nets. A computer brain, on the other hand, never gets tired or bored, so this type of processing is ideal for the implementation of autonomous vehicles using image recognition.

When the computer brain is better than the human brain

Whether the images are judged by humans or machines, we still need to make video films of the sea cage nets. Navigating an ROV through a network of marine cages is technically challenging. To get clear images, the vehicle’s camera needs to get close enough without hitting the mesh wall.

Working with autonomous systems and technologies for use in remote areas, our research colleagues are investigating the types of sensor technology that could enable an ROV to determine its spatial position within a sea cage. This type of information is key to determining which part of the network is being audited at any given time.

It is important for an ROV to know exactly where it is for any type of autonomous operation. It may be required to maintain a stable position against strong currents and high seas or to cross the net wall at a fixed predetermined distance.

The result of all this research is a laser camera system called net relative navigation. With the help of two parallel laser beams illuminating the mesh wall, information about the distance and angular orientation of the ROV to the mesh is obtained. Such measurements are important so that the ROV can maintain an appropriate distance from the net wall so that the operator does not need to steer each time the net moves due to currents or wave action. There is a lot of exciting research and development going on for autonomous robotic operations, both in research centers and in industrial settings. At SINTEF, we see this as the beginning of a system development journey that will improve operations in the aquaculture sector.

dangerous close contact

SFI Exposed is a research-based innovation center operating in many fields with a variety of perspectives on the challenges faced by offshore aquaculture operations in their environments. Our colleagues investigating safety issues have discovered that there may be some overlap between events that increase the risk of fish escape and events that threaten the safety of facility workers.

When it comes to security, we believe it’s important to consider the big picture and the most effective approach is to consider all aspects during product development. If worker safety is incorporated into technology in a way that makes it impossible to cause hazards through misuse, we effectively reduce the risk of incidents that cause both fish escape and worker injury.

But because we’re talking about risk. Obviously, a ROV does not impose HSE requirements. However, HSE considerations are essential for personnel whose job it is to release ROVs into fish cages and recover them again. This task sounds simple enough, but requires a crane-mounted vessel to be juxtaposed and attached to the cage. In many cases, personnel must be deployed along the buoy ring to release the ROV; this is a job that requires a lot of hard lifting.

Movement between ship and cage is a risky undertaking at best, and it is precisely such situations where many industry-related accidents occur. Where ships and facility infrastructure are larger than sheltered areas, which are subject to great movements and accelerations due to strong currents, open seas and winds, it is vital to have systems installed that ensure safe movement between a cage and a cage. a ship at anchor.

A robot arm can do the job

To assess the risks and challenges involved, we first examined the ways in which existing crane operations were conducted. We then continued to see how new technologies could improve the situation. An example is our concept work for the ROV marine cage “Launch and Rescue”. Another is a concept involving the use of an advanced robotic arm that allows operations to be performed without any form of contact between the ship and the cage.

Imagine that the ship to which the robot arm is attached moves up and down on the waves, and the sea cage does the same step by step and at a distance. For precision work, a robot arm that can extend from the boat to the cage is all it takes to be an impressive high-tech kit. But it’s actually entirely possible.

We used model experiments to demonstrate how a robotic arm could balance the relative motions of a ship and a sea cage. We also made evaluations about the work that can be done for the commissioning of this technology and the changes that should be made in the existing facilities.

Better decision support

The risk of accidents and fish escapes may increase when sea conditions are worse than anticipated, such as in rough seas with unexpectedly strong currents or strong waves. Currently, our decisions about whether to continue operations under such conditions are often made on the basis of common sense and experience. Better-defined operational limits, in the form of guidelines for when operations can and cannot be performed, can make a valuable contribution to improving safety.

At SINTEF Ocean we are building an infrastructure to collect data from buoys, offshore vessels and facilities, and we have collected water movement and quality measurements as well as meteorological data as a way to build a robust statistical database. The more precise the information we have about waves, currents and weather, the easier it will be to decide whether an operation should be performed. When we correlate these statistics with physical metrics and digital twins, we get an excellent basis for operational planning.

Provided by the Norwegian University of Science and Technology

Quotation: Deep learning can help prevent salmon escapes in rough seas, retrieved from https://phys.org/news/2022-12-deep-salmon-rough-seas.html on December 31, 2022 (2022, December 30)

This document is subject to copyright. No part may be reproduced without written permission, except in fair trade for personal study or research purposes. The content is for informational purposes only.

Leave a Reply

Your email address will not be published. Required fields are marked *