Line of action:

Status:

Execution date:

General Objective:

Develop and implement innovative tools using artificial intelligence techniques, such as Machine Learning/Deep Learning, and decomposition into singular values, for the identification and online quantification of the catch in a robust and reliable way, even when the specimens on the conveyor belt are overlapping and/or overlapping.

The innovative approach of the project will allow real-time decision-making through the use of maps that describe the state of fish stocks based on different criteria: fishing suitability index, presence of juveniles/vulnerable species or cost/benefit ratio. These maps are generated in a Geoportal by combining advanced GIS tools, mathematical models and data provided by the iObserver. Shipowners, scientists and/or legislators will be able to use these innovative tools to identify optimal fishing areas, avoiding unwanted catches and minimizing fishing discards, or for the development of an efficient regulation that guarantees the sustainability of fishing resources and compliance with the landing obligation.

Specific objectives:

1. Provide the iObserver system with reliable tools for species recognition with overlapping specimens
2. Improve and expand the catalogue of species included in the iObserver
3. Provide the iObserver system with complete autonomy in its operation
4. Achieve a lighting system that allows homogeneous light conditions to be obtained throughout the photograph
5. Equip the iObserver with a sensor system that allows automatic capture of photographs avoiding overlapping of images
6. Testing the efficiency of iObserver upgrades on board fishing vessels
7. Improvement of the available GIS tools and their introduction in the management of daily fishing activity in order to minimise discards and comply with the landing obligation

The SICAPTOR project, led by the IIM-CSIC entity, was developed with the support of other partner entities such as the IEO, CESGA AND OPROMAR.

The results have been validated during the 27 days of the campaign on the trawlers Portosano and Rica de Marín, in which 56 fishing sets have been carried out in real conditions on board, giving promising results and without any record of technical or operational failures of the device during them.

The project has made it possible to obtain a substantial improvement of the iObserver electronic monitoring system for the quantification of the total catch, including the discards, both of the software and the hardware, of the system through:

• The development of new species recognition algorithms, using deep learning techniques and convolutional neural networks and obtaining more than 98% accuracy in species identification and an average absolute error of 3% for size estimation (with a convolutional network created from scratch) for cases of specimens with slight or moderate overlap.
• The expansion of the catalogue to 10 species of interest to be identified by the iObserver: the new algorithms developed make it possible to identify what was initially grouped as blondes (Trigla lyra; Aspitrigla cuculus; Eutrigla gurnardus and Chelidonichthys lucerna) and roosters (Lepidorhombus boscii and Lepidorhombus whiffiagonis). In addition, the pout (Trisopterus luscus), a species of commercial interest, three species of rays (common ray –Raja clavata-; painted ray –Raja montagui-; Santiago ray –Leucoraja naevus-) of great interest in survival studies were added.
• The automation of the system, through improvements in the iObserver user software, and the development of a software solution to adequately detect the starts/ends of casts from the boat’s GPS data.
• The improvement of the lighting system through an update based on four dimmable LED spotlights with polarizing filters to avoid the appearance of areas with brightness and/or shadows that make it difficult to recognize species, improving the sharpness and resolution of the photographs obtained.
• The improvement of the sensors for image capture, avoiding overestimations/duplications in the quantification of the capture. Two solutions were implemented: on the one hand, a set of highly resistant physical sensors in aggressive marine environments, and on the other, an optical flow software algorithm that allows automatic capture adapted to different belt speeds.

It should be noted that thanks to iObserver, robust, continuous and real-time data can be obtained from the total catches that fleets are making, which feeds into the catch geoportal and the prediction map generation tools integrated into it. These maps allow for more sustainable management of daily fishing activity based on different criteria, such as the maximum percentage of discards, the presence of juveniles/vulnerable species or the fuel/benefit cost ratio. These tools have been tested on board the commercial vessels participating in SICAPTOR, obtaining very good results in terms of reducing the volume of discards, with 50-60% less biomass discarded when skippers use these GIS and prediction tools.

The products delivered in the project were the following:

• Report on software and hardware improvements implemented in the iObserver
• Results report on the use of iObserver and GIS tools on board oceanographic and commercial vessels
• SICAPTOR project factsheet on paper and digital
• Poster Results Day The future of remote electronic monitoring (REM) in fishing activity
• Poster Geographic Information Systems and Tools for Electronic Monitoring Applied to the Fisheries Sector, from the workshop aimed at the administration
• Poster for the Sicaptor Conference – iSEAS The challenge of reducing fishing discards and the landing obligation
• Sicaptor panels for IMM and OPRAMAR
• Sicaptor Project Roll-up
• Sicaptor brochures with the description of the project and the results
• 250 folders and 250 pens to distribute information during organized events
• An article in the scientific journal Fisheries Journal: Environmental implication of discarding fish in northern Spanish coastal bottom otter trawl fisheries (Annex XII)
• Two articles in the specialized magazines Industrias Pesqueras and Europa Azul: SICAPTOR, smart fishing and Remote electronic monitoring: the ‘Big Brother’ of which the fishing sector is suspicious, respectively.
• Capture geoportal: http://iseas.cesga.es
• Database containing different agents of the fisheries sector (producer associations, administrations, technology supplier/developer companies, etc.), updated during the execution of the project.
• 4 videos posted on YouTube.