A unity-based farmed fish target tracking model training data generation method

By simulating the real movement behavior of fish in a virtual environment, a highly diverse and accurate training dataset is generated, solving the problems of high data acquisition costs and large annotation workload in intelligent monitoring of aquaculture, and realizing the development of a low-cost and high-efficiency multi-target fish tracking algorithm.

CN122049571BActive Publication Date: 2026-07-24CHINA AGRI UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
CHINA AGRI UNIV
Filing Date
2026-04-16
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

In intelligent monitoring of aquaculture, existing technologies suffer from high data acquisition costs, large annotation workload, insufficient data volume, and poor scene controllability, which restricts the development and optimization of multi-target tracking algorithms for fish.

Method used

This system simulates the real movement behavior of fish in a virtual environment, reproducing the swimming and interaction scenes of fish in the water through 3D simulation technology. It uses an automatic annotation mechanism to generate annotated datasets, including fish model simulation modeling, simulated water environment construction, fish behavior state modeling, FishBrain behavior control module, FishMove motion execution module, FishSpawner batch generation and ID binding module, and MOTRecorder data recording and export module, to achieve automatic recording and output of image sequences and trajectory label data.

Benefits of technology

It reduced the cost of manual annotation, generated highly diverse training data, increased the amount and accuracy of data, reduced equipment and site construction costs, and improved the development efficiency of intelligent aquaculture systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122049571B_ABST
    Figure CN122049571B_ABST
Patent Text Reader

Abstract

The present application relates to the field of intelligent fishery and computer vision, and in particular to a method for generating training data of a target tracking model for cultured fish based on Unity, which technical solution comprises: a dataset generated through virtual simulation, which avoids high cost of manual annotation and reduces the research and development cost of the culture industry, thereby reducing the annotation cost; by adjusting parameters such as water, light, and fish behavior, training data with high diversity is generated to meet the needs of different algorithms, thereby improving data diversity; the large-scale dataset generated can be used for training and verification of algorithms, improving the development efficiency of intelligent culture systems, thereby accelerating algorithm development; fish modeling does not depend on real shooting environment, saving the cost of equipment and site construction, thereby reducing environmental dependence; has the advantages of low cost, high efficiency, and data volume can be expanded on demand, etc., can significantly reduce the difficulty of obtaining multi-target tracking training data, and ensure the accuracy and consistency of annotation.
Need to check novelty before this filing date? Find Prior Art