摘要
借助计算机视觉技术研究鱼类游泳行为已逐渐成为热点课题,它模拟生物视觉原理,通过处理采集的图片或视频获得动态目标参数信息,以达到对鱼类游泳行为监测分析的目的,本研究旨在介绍国内外该领域的研究进展,并展望其发展趋势。首先介绍鱼体监测目标的种类选择与影像获取方法,然后介绍影像中的背景去除与目标检测,并对影像数据直接和拟合提取目标参数的单个鱼体目标游泳参数提取方法,以及运动预测法和特征匹配法的多鱼体目标监测方法进行详细阐述,对游泳行为监测参数进行分类,并介绍了相关影像处理常用软件,最后总结了计算机视觉监测存在的难点及未来发展趋势。
Currently,computer vision technology which simulates the principle of biological vision has been widely applied in study of fish swimming behavior via collection and processing dynamic target parameter information including pictures or videos. Research progress on fish swimming behavior monitoring by computer vision technology is summarized. Firstly,the fish target selection and image acquisition methods are introduced. Secondly the background removal and target detection are presented. Thirdly extraction method of swimming parameters for single fish target with direct extraction and fitting extraction is presented. The method of motion prediction and feature matching for multi fish object monitoring is described in detail. Then the swimming behavior monitoring parameters are classified,and relevant image processing software used commonly is introduced. Finally,the difficulties and future trends of computer vision monitoring are discussed.
出处
《大连海洋大学学报》
CAS
CSCD
北大核心
2017年第4期493-500,共8页
Journal of Dalian Ocean University
基金
中央级公益性科研院所基本科研业务费专项(东海水产研究所2016T01)
上海市自然科学基金资助项目(15ZR1450000)
河口海岸学国家重点实验室开放基金资助项目(SKLEC-KF201403)
关键词
计算机视觉
游泳行为
目标检测
鱼类跟踪
computer vision
swimming behavior
target detection
fish tracking