基于3D计算机视觉的鱼类行为分析研究

Research on fish behavior analysis based on 3D computer vision

分析了现有鱼类生物监测系统不能全面精准地检测和描述鱼类的空间位置的缺陷,设计了一种基于3D计算机视觉的鱼类行为分析系统,以实现从三维视角对鱼类进行监测并分析鱼类的运动行为。该系统首先通过自组织背景建模方法对图像中鱼的前景进行提取,然后计算出鱼的质心位置,并由3D视觉装置的几何关系计算鱼在实际三维空间中的位置,通过跟踪得到鱼类运动的三维轨迹,最后根据鱼类运动轨迹数据计算出鱼的游动的速度和角速度,得到鱼类运动行为模式曲线。实验结果表明利用3D计算机视觉对鱼类行为进行跟踪与描述,能为鱼类活动量水平变化及鱼类活动模式改变的定量分析奠定坚实的基础。

Considering that existing fish biological monitoring systems can not comprehensively detect and describe the movement behavior of fish, a new kind of fish behavior analysis system based on 3D computer vision was designed to realize the 3D fish monitoring and 3D fish behavior analysis. The system can fully monitor and detect fish move- ment behaviors. It extracts the foreground image of fish by using the self-organization background modeling method and calculates the centroid position of fish by using the experiment device of geometric relations. Then it can get the fish movement 3D trajectory. According to the trajectory data, the fish swimming velocity and angular velocity can be calculated. The experimental result shows that it can be the solid foundation of later research to take advantage of 3D computer vision to track and describe fish behaviors.