An Integrated Framework for Shrimp Detection, Tracking, Counting, and Behavioral Analysis Using YOLOv8 and Deep Learning-Based Computer Vision Model

Artificial intelligence-based machine learning (ML), deep learning (DL) models are used to detect, track, and count shrimp underwater, which can be used to monitor the health and behavior of the shrimp, optimize the water quality, and identify problems early on. This study investigated the use of convolutional neural networks (CNNs), YOLOv5 and YOLOv8, ML, and DL models for underwater shrimp detection, tracking, and counting in glass tanks. A custom-built near-infrared night vision camera system was used to capture images and videos of shrimp in glass tanks. A deep learning algorithm was then developed using CNNs, YOLOv5 and YOLOv8, ML, and DL to detect, track, and count the shrimps in different environmental conditions. The tracking and counting data were used to analyze the behavior of the shrimp by different ML classifiers (decision tree, naïve Bayes classifier, and linear discriminant analysis (LDA)) with three attributes (water temperature, pH, and dissolved oxygen) to observe changes in swimming pattern and feeding ability, etc. The results showed that the shrimp exhibited different behaviors depending on the water quality changes for temperature, DO, and pH. This ML and DL technologies can revolutionize the shrimp farming industry and help ensure a more sustainable future for shrimp production.