毛家发，胡海彪，程振波，范伟康，陆佳炜，肖　刚

(浙江工业大学 计算机科学与技术学院，浙江 杭州 310023)

半遮挡目标鱼体的识别与跟踪方法研究

毛家发，胡海彪，程振波，范伟康，陆佳炜，肖刚

(浙江工业大学 计算机科学与技术学院，浙江 杭州 310023)

摘要：基于平面镜原理的三维目标跟踪平台，半遮挡会引起多目标交叉连接在一起，难以清晰地将多目标独立地分离出来，降低了跟踪定位准确性.针对这一问题，提出一种基于计算机视觉的半遮挡鱼体识别的方法.通过鱼体轮廓识别技术和鱼体质心点的提取方法，解决水箱当中鱼体目标与镜子当中鱼体目标的对应关系.最后进行了活体鱼半遮挡实验，实验结果表明：该方法能够有效地解决鱼体半遮挡分离问题，提高了鱼目标跟踪定位的准确性。

关键词：半遮挡；鱼体轮廓；鱼体识别；鱼体质心

Study on identification and tracking of half-occlusion target fish

MAO Jiafa, HU Haibiao, CHENG Zhenbo, FAN Weikang, LU Jiawei, XIAO Gang

(College of Computer Science and Technology, Zhejiang University of Technology, Hangzhou 310023, China)

Abstract:For the three-dimensional object tracking platform based on principle of plane mirror, half-occluded objects can cause multi-objective overlap. Thus it is difficult to clearly separate the targets and tracking accuracy will be reduced. A half-occluded fish contour recognition method based on computer vision is proposed. Using fish contour recognition technology and fish centroid point extraction method, we have obtained the coordinate spatial relationships between the target fish in tank and virtual fish in the mirror. Here we test the real fish in the experiment and experimental results show that this method can effectively solve the problem of separating half-occluded fish and the fish target tracking accuracy is improved。

Keywords:half-occluded; fish contour; fish recognition; fish centroid

鱼类作为水质监测的重要指示生物，很多学者通过研究其运动行为来反应水质情况.鱼类的主要行为特征为游动轨迹、速度、角速度、加速度等[1].如果能够通过图像处理技术对鱼运动视频进行跟踪处理，提取鱼目标的运动参数，那么便能采用仿真鱼进行水下模拟真实鱼的生理行为并且进行水下作业，进而为军事和民用研究提供更好的实际依据[2].水环境的变化会明显地影响鱼类的生活习性，因此可以通过提取鱼类的行为特征参数，来反应水环境变化情况，鱼类行为数据的变化为水质监测提供重要……