基于视频算法的鱼类运动跟踪研究  

Fish Motion Tracking Research Based on Video Algorithm

作　　者：程淑红[1] 蔡菁[1] 胡春海[1] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《光电工程》2011年第2期14-18,共5页Opto-Electronic Engineering

基　　金：河北省教育厅发展计划基金资助项目(2006455)

摘　　要：针对利用鱼类行为监测进行水体环境保护的问题,本文提出了基于视频的鱼类运动跟踪研究,通过对鱼类运动视频进行分割、跟踪,获得鱼类运动的轨迹和速度,为鱼类参与环境污染研究奠定了理论基础(通过对比不同污染环境中鱼类运动的一些参数,进行水体环境污染程度的定量分析)。该算法采用标记多尺度分水岭方法进行鱼类运动分割,然后通过改进的加权Hausdorff距离对鱼类运动视频进行跟踪,最后为了容纳鱼类在运动过程中形状的变化,在多值图像中引入欧几里德范数作为约束条件来完成跟踪模型的更新。实验结果表明,本文算法呈稳定跟踪状态,在连续100帧的跟踪过程中没有出现超过1个像素的位置差,跟踪速度差值也未超过0.12个像素,能够快速、精确分割和跟踪鱼类运动目标。Aiming at the problem of water environment protection using fish motion tracing, the fish motion monitor research is proposed based on the video algorithm. It can obtain trajectory and velocity of the fish by means of segmentation and tracing of the fish motion video, which lie a theory foundation for the fish to participate in the environment pollution research （through the comparison of some parameters of the fish motion in different polluted environment, to carry out quantitative analysis of the water pollution degree）. The algorithm uses marker multi-scale watershed method for fish motion segmentation, then tracks the fish motion video with the improved weighted-Hausdorff. Finally, in order to accommodate the shape change of the fish in the moving process, introduce the Euclid norm into multi-value images as the restriction to accomplish the update of the tracing model. The experiment shows that the proposed method presents the state of stable tracking, the position shift in the tracking of the successive 100 frame is within a pixel, and the tracing velocity difference is within 0.12 pixels, which can segment and track the fish motion object fleetly and precisely.

关 键 词：视频算法 鱼类 分割 跟踪:环境保护 

分 类 号：TN919.8[电子电信—通信与信息系统]