Girona Vision Research
Leader in computer vision and machine learning for the ocean industries
Girona Vision Research (GVR)
Girona Vision Research is a company working of the development of software for the image acquisition and analysis, specially focused to the underwater context.
The company carries out both research and development tasks to implement tailored solutions based on computer vision and machine learning techniques. We are experts on methods for the acquisition and manipulation of images acquired in the underwater environment, which require a specific processing, more complex than the one applied on conventional on-air images.
The activity of the company is mainly focused on the processing of images acquired by the Scantrol Deep Vision system. The main goal of this analysis is the gather of statistics allowing to describe its population. Among the tasks being done using this imagery, the most remarkable are the identification and classification of the individuals, the estimation of its size and the counting of the individuals being imaged.
Almost 90% of global commercial fish stocks are overfished or fully exploited and 1/3 of the stocks are overfished. At the same time, we need to feed a growing population reaching 9.8 billion by 2050, and global demand for seafood is expected to double. Today's fishing methods are inefficient and unsustainable, losing approximately $83 billion a year in revenue. In order to make changes in ocean harvesting, we need to know what kind of fish and species enter the trawl before it lands on deck.
In close collaboration with its sister company Scantrol Deep Vision, Girona Vision Research (GVR) is able to address these challenges through the development of vital computer vision, machine learning, and autonomous control technologies.
To improve catch efficiency and reduce bycatch to a minimum, the main goal is to deliver realtime fish sampling and analysis in the trawl to the global fishing industry.
Key Challenge
The Solution
Deep Vision Image Processing Technology
Partnering with Scantrol Deep Vision, a complete acquisition and processing pipeline has been developed for efficient catch identification in the trawl, successfully tested, and used in real scenarios for marine research purposes.
Beyond acoustics imagery, optical imaging goes one step further in precision and automatic sampling, identification, and measurements reducing bycatch, optimizing catch value, and contributing to improved biodiversity.
Key Functionality
Color correction provides image appearance uniformization among acquisition units, as well as accurate color analysis. The catch pictures look as if they had been taken out of the water. This is vital functionality for fish identification below water.
Species distribution assessment is based on specimen identification, as well as automatic measurement and eventual weight estimation that can be performed thanks to optical imaging analysis
Optical image acquisition and analysis at all commercial trawling depths demonstrates unmatched capabilities. The below images demonstrate the process from the original fish image to a segmented fish below.
Original Image
Image after colour correction
Segmented fish
Key Benefits
Precise individual identification enables species behavior analysis, allowing the application of optimized strategies to improve catching performance and bycatch reduction.
Access to the fishing industry to gather datasets creates clear benefits. The data transparently collected by hundreds of ships will allow for improving the precision and efficiency of the acquisition system, as well as develop solutions for additional challenges, intended to optimize catch, reducing time and resources.
Ultimately the developed technology and solution allow automatic fish measurement without a physical catch. The developed technology enables deep learning solutions for real-time species classification and consequent population assessment.
Deep Vision & Girona Vision Research
Scantrol's sister companies Deep Vision and Girona Vision Research are key partners in our vision & control strategy. The two companies are experts in vision technology and machine learning and work together to develop an autonomous subsea vision system that allows commercial fishermen to catch only legal fish while preventing the catch of illegal or undersized fish. The system is already in use within marine research, and commercial fishermen will soon be able to use the system.
Marine resources can be harvested sustainably through the introduction of new technologies and a change in attitudes. We aim to take part in this development by providing vision & control technology for marine research and commercial fisheries.
The above video shows the Deep Vision underwater camera system in action taking images of cod swimming through the trawl and corresponding screenshots.