Abstract:Feed has the greatest impact on the carbon footprint of the aquaculture, and also determines the water quality in aquaculture to a great extent. Making appropriate feeding control strategies is one of the most effective ways to promote cleaner production as well as fish welfare in aquaculture. Reliable and accurate fish appetite grading especially based on time-limited data is a prerequisite for achieving high-precision and reasonable feeding control in practical production. To date, however, few efforts have been done on this challenge. For these, regarding Micropterus salmoides as the experimental fish, a novel and practical method, based on a modified kinetic energy feature-based graph convolutional network (GCN), was developed in this study. First, graphs were constructed based on the extracted modified kinetic energy features and their temporal correlation. Then, with the help of a series of the convolution and global pooling operations, a GCN model was customized based on the constructed graphs. Following this, the customized GCN model was enriched by the self-attention pooling mechanism and customized network structure. Results show that the proposed GCN-based approach outperforms other typical state-of-the-art methods in fish appetite grading, and the grading accuracy obtained here could be 98.60% using only the first 4.2 seconds as well as the first 8.3 seconds of input data, which is not much different from that (98.89%) using full-length (25 second-long) input data. What’s more, compared to the recurrent neural network (RNN)-based method which performance is closest to our method, the space complexity of the proposed approach here can better satisfy the requirements of real aquaculture, in which the quantity of the trainable parameters here is only 6.4% ~ 31.8% of the RNN-based method. In summary, the proposed modified kinetic energy feature-based GCN approach is favorable for the appetite grading of fish like Micropterus salmoides with time-limited data, which is a promising approach in dealing with feeding control tasks and alleviating the water environmental burden in aquaculture.

A Classification Algorithm of Fish Feeding Behavior for Automatic Bait Feeding Control

Multi-class Fish Stock Statistics Technology Based on Object Classification and Tracking Algorithm

Modified Kinetic Energy Feature-Based Graph Convolutional Network for Fish Appetite Grading Using Time-Limited Data in Aquaculture

Behavioral spatial-temporal characteristics-based appetite assessment for fish school in recirculating aquaculture systems

Behavioral Characteristics and Statistics-Based Imaging Techniques in the Assessment and Optimization of Tilapia Feeding in a Recirculating Aquaculture System

A real-time feeding decision method based on density estimation of farmed fish

Modified Motion Influence Map and Recurrent Neural Network-Based Monitoring of the Local Unusual Behaviors for Fish School in Intensive Aquaculture.

Fish Detection and Classification Based on Improved ViT

A hybrid method for identifying the feeding behavior of tilapia

A Robust Underwater Multiclass Fish-School Tracking Algorithm

Dynamic Scene Images-Assisted Intelligent Control Method for Industrialized Feeding Through Deep Vision Learning.

A Deep Vision Sensing‐based Fuzzy Control Scheme for Smart Feeding in the Industrial Recirculating Aquaculture Systems

Research on intelligent bait casting method based on machine vision technology

Automatic Controlling Fish Feeding Machine using Feature Extraction of Nutriment and Ripple Behavior

A Computer Vision-Based Intelligent Fish Feeding System Using Deep Learning Techniques for Aquaculture

A recurrent network based on active learning for the assessment of fish feeding status

Research on feeding behaviour of fish by using spatial and temporal features of depth images

An integrated three-stream network model for discriminating fish feeding intensity using multi-feature analysis and deep learning

A high-density fish school segmentation framework for biomass statistics in a deep-sea cage

An Automated Fish-Feeding System Based on CNN and GRU Neural Networks

Fish Trajectory Extraction Based on Object Detection