Edge computing based real-time Nephrops (Nephrops norvegicus) catch estimation in demersal trawls using object detection models.

In demersal trawl fisheries, the unavailability of the catch information until the end of the catching process is a drawback, leading to seabed impacts, bycatches and reducing the economic performance of the fisheries. The emergence of in-trawl cameras to observe catches in real-time can provide such information. This data needs to be processed in real-time to determine the catch compositions and rates, eventually improving sustainability and economic performance of the fisheries. In this study, a real-time underwater video processing system counting the Nephrops individuals entering the trawl has been developed using object detection and tracking methods on an edge device (NVIDIA Jetson AGX Orin). Seven state-of-the-art YOLO models were tested to discover the appropriate training settings and YOLO model. To achieve real-time processing and accurate counting simultaneously, four frame skipping ideas were evaluated. It has been shown that adaptive frame skipping approach, together with YOLOv8s model, can increase the processing speed up to 97.47 FPS while achieving correct count rate and F-score of 82.57% and 0.86, respectively. In conclusion, this system can improve the sustainability of the Nephrops directed trawl fishery by providing catch information in real-time.

Towards sustainable demersal fisheries: NepCon image acquisition system for automatic Nephrops norvegicus detection.

Underwater video monitoring systems are being widely used in fisheries to investigate fish behavior in relation to fishing gear and fishing gear performance during fishing. Such systems can be useful to evaluate the catch composition as well. In demersal trawl fisheries, however, their applicability can be challenged by low light conditions, mobilized sediment and scattering in murky waters. In this study, we introduce a novel observation system (called NepCon) which aims at reducing current limitations by combining an optimized image acquisition setup and tailored image analyses software. The NepCon system includes a high-contrast background to enhance the visibility of the target objects, a compact camera and an artificial light source. The image analysis software includes a machine learning algorithm which is evaluated here to test automatic detection and count of Norway lobster (Nephrops norvegicus). NepCon is specifically designed for applications in demersal trawls and this first phase aims at increasing the accuracy of N. norvegicus detection at the data acquisition level. To find the best contrasting background for the purpose we compared the output of four image segmentation methods applied to static images of N. norvegicus fixed in front of four test background colors. The background color with the best performance was then used to evaluate computer vision and deep learning approaches for automatic detection, tracking and counting of N. norvegicus in the videos. In this initial phase we tested the system in an experimental setting to understand the feasibility of the system for future implementation in real demersal fishing conditions. The N. norvegicus directed trawl fishery typically has no assistance from underwater observation technology and therefore are largely conducted blindly. The demonstrated perception system achieves 76% accuracy (F-score) in automatic detection and count of N. norvegicus, which provides a significant elevation of the current benchmark.

When is enough, enough? Quantifying trade-offs between information quality and sampling effort for fishing gear selectivity data.

There is general pressure throughout the world's fisheries for the industry to have greater involvement not only in the development of fishing gears but also in the testing and documentation of their effect. In the European Union, the Common Fisheries Policy of 2013, together with the proposed reform of the technical measures regulation, highlights the need for greater flexibility in fisheries through increased stakeholder involvement. To achieve this flexibility, there is a need for additional fishing gears available to the fishermen. A way to facilitate this is to have the industry take part in the development and testing of fishing gears, as well as collect data on their performance. However, to have a successful industry-collected data programme, fishermen have to be able to collect data on the length of a portion of the catch. In this study, we determine how many individuals need to be measured to correctly evaluate the relative selective performance of a new gear compared to a standard gear. The evaluation was carried out by analysing catch ratio curves, their associated uncertainties, and the trade-offs between uncertainties and sampling effort. Results show that with relatively small sample sizes (500 to 1000 individuals) it is possible to correctly evaluate the performance of a gear for a given species. By having the industry develop and test their own gears, as well as being involved in the collection of data, the number of potential gear solutions available to address the different issues emerging in the fisheries is increased.

From Fishing to Fish Processing: Separation of Fish from Crustaceans in the Norway Lobster-Directed Multispecies Trawl Fishery Improves Seafood Quality.

Fishing gears have negative impacts on seafood quality, especially on fish in the mixed trawl fishery targeting Norway lobster (Nephrops norvegicus). In this fishery, which is worth about €80 millions in Denmark alone, the quality of fish can be significantly improved by simple gear changes. A trawl codend divided into an upper and lower codend was designed to separate fish from Norway lobster during the fishing process by encourage fish to swim into the upper codend by using a frame at the entrance of the lower codend. Separate codends for fish and Norway lobster in the same gear provide the opportunity to selectively reduce small low-value fish, which will reduce catch weight and sorting time onboard the vessel. For this horizontally divided test codend and a standard codend, in which the catch was mixed, quality assessments were performed on the same batches of fish during three steps of the value chain: i) aboard the fishing vessel; ii) at the Fishermen's Collection Central, and iii) in the production plant. Four species of fish and fillets from fish caught in the upper codend of the test codend were of significantly better quality for several of the assessed parameters compared with those caught in the standard codend: i) newly caught fish showed significantly less scale loss and discolourations and had significantly better texture; ii) landed fish had significantly better skin appearance and texture and significantly fewer discolourations; and iii) fillets showed significantly fewer blood spots and had significantly better texture. There were no differences in injuries for newly caught fish or gaping and bruises for fillets between the test and standard codends. The decrease in catch-related damages in the test codend is explained by little contact between fish and animals with hard or spiny surfaces due to successful separation of fish and Norway lobster into the upper and lower codends, respectively, and by lower catch weight in the upper codend of the test codend compared with the standard codend. The decrease in damages may also improve quality indirectly by inflicting less stress to the fish and subsequently give better texture, which offers advantages such as pre-rigor filleting and fresher products for the market. Significant improvements in fish quality can potentially increase the catch value in nationally important fisheries.

Inferring fish escape behaviour in trawls based on catch comparison data: model development and evaluation based on data from Skagerrak, Denmark.

During the fishing process, fish react to a trawl with a series of behaviours that often are species and size specific. Thus, a thorough understanding of fish behaviour in relation to fishing gear and a scientific understanding of the ability of different gear designs to utilize or stimulate various behavioural patterns during the catching process are essential for developing more efficient, selective, and environmentally friendly trawls. Although many behavioural studies using optical and acoustic observation systems have been conducted, harsh observation conditions on the fishing grounds often hamper the ability to directly observe fish behaviour in relation to fishing gear. As an alternative to optical and acoustic methods, we developed and applied a new mathematical model to catch data to extract detailed and quantitative information about species- and size-dependent escape behaviour in towed fishing gear such as trawls. We used catch comparison data collected with a twin trawl setup; the only difference between the two trawls was that a 12 m long upper section was replaced with 800 mm diamond meshes in one of them. We investigated the length-based escape behaviour of cod (Gadus morhua), haddock (Melanogrammus aeglefinus), saithe (Pollachius virens), witch flounder (Glyptocephalus cynoglossus), and lemon sole (Microstomus kitt) and quantified the extent to which behavioural responses set limits for the large mesh panel's selective efficiency. Around 85% of saithe, 80% of haddock, 44% of witch flounder, 55% of lemon sole, and 55% of cod (below 68 cm) contacted the large mesh panel and escaped. We also demonstrated the need to account for potential selectivity in the trawl body, as it can bias the assessment of length-based escape behaviour. Our indirect assessment of fish behaviour was in agreement with the direct observations made for the same species in a similar section of the trawl body reported in the literature.