基于边缘重心模板的水中机器鱼阈值分割算法研究

Research on Threshold Segmentation Algorithm of Underwater Robotic Fish Based on Edge Gravity Center Template

为了实现不同光照条件下水中机器鱼的准确视觉分割,提出了一种基于边缘重心模板的阈值分割新算法。该算法首先将机器鱼图像从RGB颜色空间转化至HSV颜色空间,利用图像H分量直方图来确定机器鱼像素H分量值的极差;然后基于Sobel算子进行边缘提取来获取边缘重心坐标,并计算以边缘重心为中心的n×n模板的H分量平均值;最后利用机器鱼像素H分量值的极差和平均值来确定边缘重心模板阈值,从而实现机器鱼的阈值分割。3种光照环境的实验结果表明,与其他算法相比,基于边缘重心模板的阈值分割法不仅保证了平均96%的高检出率,其平均误检率也控制在0.4%,从而验证了此算法的有效性。

To achieve accurate visual segmentation of an underwater robotic fish in different lighting conditions,this paper proposes a novel segmentation algorithm based on edge gravity center template.Firstly,the image of robotic fish was converted from RGB color space to HSV color space,and the H-component histogram was used to determine the extreme difference of the Hcomponent value of the robotic fish.Then,the edge center coordinates were obtained through the Sobel operator-based edge extraction,and the average H-component value of n×nedge gravity center template was calculated.Finally,the extreme difference and average value of H-component were used to determine the threshold value of edge gravity center template,and thus the robotic fish was segmented.The experimental results of three kinds of lighting conditions show that,compared with other algorithms,the proposed algorithm not only guarantees the high average detection rate of 96%,but also controls the average false detection rate at 0.4%.