This invention concerns an intelligent
system for
fishing selectivity and catch optimization of coastal fishermen designed to improve
fishing efficiency in an
environmentally friendly manner. This
system consists of an
environmentally friendly fish trap equipped with an
underwater camera, aquatic environmental sensors, an
underwater lighting module, an
edge computing-based
microcomputer unit, a communication module, and a
power management module. Fish images and
environmental data are automatically processed using the EcoTrapNet-DNN
algorithm, which is a multi-task deep neural network based on
underwater acquisition parameters configured to perform
image quality validation,
image quality improvement, fish detection and segmentation, size
estimation based on geometry calibration, classification of fish types and sizes, and determination of catch selectivity scores.Based on the analysis, the
system provides recommendations for
fishing selectivity or, in certain embodiments, controls the selective mechanism in the
fish trap so that target fish are retained, while non-target or substandard fish are released. This system can reduce bycatch, improve the quality and value of the catch, and support sustainable fishing practices for small- to medium-scale coastal fishers.