Abstract:Currently, aquaculture methods tend to combine scale and intelligence, which saves manpower and improves the survival rate of seafood at the same time. High-precision and high-efficiency fish individual recognition can provide key technical support for fish disease detection, feeding habits, body condition, etc. In the realm of intelligent aquaculture, it provides robust data support for precision fish farming. However, the current research methods for individual fish recognition struggle to maintain the network model's focus on the fish body in real marine underwater complex environments (e.g., environmental background interference such as coral reefs, overlap between fish bodies, light noise, etc.), leading to unsatisfactory recognition results. To this end, this paper proposes a method for fish individual recognition in underwater complex environments based on video object segmentation, which consists of three parts, including a fish individual segmentation detection module, a fish individual recognition module, and an all-in-one visualization module. The work adopts a combination of deep learning methods and video object segmentation algorithms to solve the problem of low attention and poor detection accuracy of fish individuals in real underwater complex environments, which effectively improves the accuracy and efficiency of fish individual recognition, and analyzes and discusses the comparison of recognition effects using different weights. The results of the simulation experiments show that the key metric Rank1 value of the method achieves more than 96% accuracy on the public datasets DlouFish, WideFish, and the Fish-seg dataset produced in this paper, and improves over the state-of-the-art methods for fish individual recognition by 2.23%, 1.33%, and 1.25%, respectively.

Research on multi-object recognition algorithm based on video

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

Embedded Online Fish Detection and Tracking System Via YOLOv3 and Parallel Correlation Filter

A Robust Underwater Multiclass Fish-School Tracking Algorithm

Multi-objects Real Time Recognition Based on Color Information

MDM-YOLO: Research on Object Detection Algorithm Based on Improved YOLOv4 for Marine Organisms.

An Automatic Detection and Statistical Method for Underwater Fish Based on Foreground Region Convolution Network (FR-CNN)

YoloXT: A Object Detection Algorithm for Marine Benthos

Research on Underwater Small Target Detection Algorithm Based on Improved YOLOv3

Fish Detection and Classification Based on Improved ViT

Improved Objects Matching with a Single RGB Camera

Fish Trajectory Extraction Based on Object Detection

A video object segmentation-based fish individual recognition method for underwater complex environments

Underwater Object Detection Based on Enhanced YOLO

A Fish Detection and Tracking Method Based on Improved Interframe Difference and YOLO-CTS

An Image Segmentation Algorithm Based on Edge Detection and Its Application to Robotfish

Research on underwater object recognition based on YOLOv3

An Effective and Robust Method for Tracking Multiple Fish in Video Image Based on Fish Head Detection

Marine Fish Object Detection Based on YOLOv5 and Attention Mechanism.

Research on multi-scale fusion image enhancement and improved YOLOv5s lightweight ROV underwater target detection method

A multitask model for realtime fish detection and segmentation based on YOLOv5