Abstract.
In Chile, The National Fisheries and Aquaculture Service (SERNAPESCA) supervises, controls, and registers the extraction and compliance with the catch quotas of the fisheries at the national level. The anchovy and common sardine fish species correspond to 70.9% of the country's total quota of pelagic artisanal fishing resources, and the highest quotas are assigned to the Biobío and Ñuble Regions. The Chilean regulation states that all artisanal fishing landings must be inspected on the spot. However, the inspection process is not easy because the number of inspectors is limited, the number of artisanal vessels is large, the volumes of fish extracted are massive, and the number of landings of each artisanal shipowner is high.
The Chilean regulation associated with the inspection process is dated, manual, and takes at least three samples, using 3-liter buckets from each landing of each artisanal shipowner. To control and register the catch quotas, officers classify the species by expert knowledge, and then, each set of species is weighted to estimate the fish distribution of the landing. Then, such data is written by hand into paper spreadsheets. Further, the control process does not store evidence of the samples, so this exercise is not traceable. The uncertainty in the control operations, which are an input to determine future catch quotas, harms the fulfillment of the strategic objectives and the mission of SERNAPESCA and, more importantly, threatens the sustainability of the resources.
To solve the problem, we will design and validate a machine to determine the fish species distribution of each landing in an industrial scenario. We have termed our solution "The artificial vision gantry for fish species discrimination." To construct the gantry, we will carry out applied research in spectral computational vision and machine learning. We will integrate hardware, software, and algorithms for discriminating small pelagic species from their by-catch species, which may or may not be subject to catch-control quotas. We will focus on artisanal fishing for sardines and anchovies in the Biobío and Ñuble Regions. This solution will be a modern fisheries inspection tool that will automatically discriminate the species of interest (common sardine and anchovy). Remarkably, by continuously monitoring fish landings, the artificial vision gantry will sample landings in a non-invasive, homogeneous, and representative manner, which is impossible for humans. Moreover, the artificial vision gantry will introduce objective fish classification, quantified uncertainty, and traceability to the catch-control process.
Funding agency: FONDEF.
Program: IX CONCURSO DE INVESTIGACION TECNOLOGICA IDeA DE FONDEF.
Grant number: IT20I0032.
Funding period: July 2021 — Jubne 2023.
Director: Jorge E. Pezoa.