Sea Cucumber Monitoring with AI Detection

As with many popular marine species, sea cucumbers have been subject to overfishing. But as the fishing industry continues to adapt, sustainability and resource management have become a global priority. This has sparked a growing interest in monitoring technologies to help detect and analyze the ongoing health of marine stock.

Supported by the National Research Council Canada (NRC) and the Atlantic Canada Opportunities Agency (ACOA), Marine Thinking developed an artificial intelligence (AI) program to automatically detect, recognize, and identify sea cucumbers along the ocean floor.

Machine Learning

To train Marine Thinking’s AI program, our team used remote operated vehicles (ROV) to navigate the ocean floor and collect hours of optical footage across multiple sessions.

Using portions of the captured footage, the team created several large data sets – manually identifying and labelling sea cucumbers for the machine learning algorithm. The training process included multiple image augmentation techniques to optimize overall learn quality. These techniques help us prevent data overfitting, increase prediction confidence, and improve the overall variability of the data models.

AI Detection, Recognition, and Identification

Once fully trained, the AI program could successfully detect, recognize, and identify sea cucumbers during real-time playback of new footage. To visualize the collected data, Marine Thinking produced maps of each fishing zone, highlighting the density of sea cucumbers found in each area along with GPS coordinates.

AI Detection – Sea Cucumbers (2020)

Working in tandem with the initial project scope, the development team would apply the same machine learning techniques and AI program to detect 20 additional marine species.

AI Detection – Additional Marine Species (2020)

Sea Cucumbers: Our AI Detection Prologue

Marine Thinking has continued to solidify its expertise in AI development, creating multiple models with advanced training techniques for object detection, recognition, and identification.

Marine Thinking’s current AI detection model is integrated into Marine Tensor™ , our AI-powered control and communication system for surface vessels. Along with sensor fusion of multiple data sets above and below the water (optical, LiDAR, infrared, radar, and sonar), the AI detection model allows Marine Tensor™ to perform with greater situational awareness when conducting autonomous navigation.

AI Detection – Vessel Navigation (2023)

To learn how Marine Thinking can integrate AI detection and autonomous navigation into your vessel, please contact our sales team.