FOTOPEIX – Integrated traceability units: technological developments for the automatic calculation of catch size from images obtained at fish landing points
Pleamar program
The project sought to develop analytical tools to determine the sizes of the fish products landed in each port of the Balearic Islands, applying an important technological component to the sustainable management of fisheries resources. The initiative has focused on the development and testing of an algorithm for extracting sizes of a species, the llampuga (Coryphaena hippurus), from standard photos that are sent from the Integrated Traceability Units (UITs), scales with an integrated automatic camera and experimentally located by the Government of the Balearic Islands in the pilot port.
The information generated by ITUs and fine-tuned with this project is based on the following process:
- Each fisherman places each unloaded box at ITU at the port of unloading.
- A weight value, species composition of the box, photo, date, time and characteristics of the boat are generated.
- This information is sent to IMEDEA (beneficiary entity) automatically.
- In IMEDEA, this information is processed, based on the algorithm developed in FOTOPEIX, so that an average size is assigned to the species of that specific box.
This information allows us to know in practically real time the variation in size of certain species and improve fisheries management systems and product traceability.
Line of action:
Marine ecosystemsStatus:
FinalizadoExecution date:
2018General Objective:
Develop the analytical tools necessary to determine the sizes of fish products landed in each port.
Specific objectives:
- Develop and test the unsupervised mathematical algorithm that extracts the fish sizes of a pilot species from images of the boxes landed in a single pilot port.
- Implement the necessary architecture to store, manage and provide the information generated.
- Disseminate the positive implications of ITU implementation for the entire fisheries sector.
The FOTOPEIX project has carried out the obtaining of images of individualized fish boxes with sizes between 20 and 80 cm, in order to acquire a database of morphometric measurements, with hake being the species selected for its significant variation in size throughout the season, as well as for its seasonality. A total of 123 hake have been measured, with a total length ranging from 19 cm to 62 cm.
The measurement of a part of the fish’s body (head) was studied in order to estimate the total size – a part is selected, since, normally, the catches are not fully visible in the boxes. Numerous images were obtained in different positions and postures with which they have tried to validate the predictions of the developed system. Different phenomenological models have been considered and all of them have shown a similar predictive capacity.
The algorithm has been 86% accurate in its predictions. A percentage of success that could be improved by increasing the number of heads in the database. An unsupervised algorithm measurement study was carried out, detecting 277 heads out of a total of 284 (97.5%). Comparing the length obtained from the image and the actual height, the residual variance was 0.5 cm. Regarding the estimation of the size of the body length, the estimates obtained do not seem to show any systematic bias, being around 5 cm for fish between 40 and 60 cm and registering a greater error in lower specimens, due to their great dependence on precision and posture.
We highlight the identification of hake heads in images of boxes in the fish markets, the advances in the estimation of their size from the size of the head and the development of an architecture for the mass storage of images and their management.
FOTOPEIX – Integrated traceability units: technological developments for the automatic calculation of catch size from images obtained at fish landing points
back to search 