基于计算机视觉技术的运动鱼检测算法  

Fish motion detecting algorithms based on computer vision technologies

作　　者：范良忠[1] 刘鹰[2] 余心杰[1] 卢焕达[1] 

机构地区：[1]浙江大学宁波理工学院,宁波315100 [2]中国科学院海洋研究所,青岛266071

出　　处：《农业工程学报》2011年第7期226-230,I0003,共6页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金(30972267);TJ2X2-YW-16

摘　　要：为了提高水中鱼的运动检测精度,该文在分析比较图像和视觉领域常用的运动目标检测算法的基础上,提出了3种适合于运动鱼检测的改进算法。试验结果表明,该文提出的算法在检测鱼的快速运动方面要优于传统算法,特别是帧间差分和运动平均背景建模结合方法较其他算法在平均检测时间和检测精度方面更具优势,在鱼类运动检测方面有较大的应用潜力。In order to improve the detecting accuracy of fish motion,based on the analysis and comparison of classical moving object detection algorithms,three improved detection methods suitable for moving fish detection were presented.Experimental results showed that the proposed fish motion detecting algorithms had advantage over traditional methods.Especially,moving average background and the temporal difference method was better than other algorithms in detecting time and detecting accuracy,it has great potential in fish motion detection.

关 键 词：鱼 行为研究 计算机视觉 运动检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]