摘要
为了克服人工观测鱼群异常行为费时费力的问题,本文提出了一种基于图像处理和压缩感知算法的鱼群低溶氧胁迫异常行为的自动检测方法。以锦鲤(Cyprinus carpio)为研究对象,通过获取常氧和低氧2种情况下的鱼群运动视频图像,利用图像处理技术得到鱼群位置直方图,提取鱼群位置的均值、方差、歪斜度、峰态和能量5个参数,构成每幅图像的鱼群运动特征参数。在此基础上构建数据词典矩阵,并利用压缩感知分类方法实现低溶氧胁迫下的鱼群异常行为检测。实验结果表明,该方法能有效实现低溶氧胁迫下的鱼群异常行为检测,准确率达到98.50%。
In order to overcome the time-consuming and laborious problems of artificial observation,we proposed an automatic detection method of abnormal behavior of fish school under low dissolved oxygen stress based on image processing and compressed sensing algorithm.Taking Cyprinus carpio as the research object,by obtaining the video images of fish school behaviors under two conditions of normoxia and hypoxia,we used the image processing technology to get the location histogram of fish school,of which the average,variance,skewness,kurtosis and energy were extracted to form the fish movement characteristic parameters of each image.On this basis,the data dictionary matrix was constructed,and the abnormal behavior detection of fish school under low dissolved oxygen stress was implemented by compressed sensing classification.The results showed that the detection method can effectively detect the abnormal behavior of fish school under the low dissolved oxygen stress,with the detection accuracy rate of 98.50%.
作者
卢焕达
于欣
刘广强
LU Huanda;YU Xin;LIU Guangqiang(Ningbo Institute of Technology,Zhejiang University,Ningbo 315100,Zhejiang,China)
出处
《浙江大学学报（农业与生命科学版）》
CAS
CSCD
北大核心
2018年第4期499-506,共8页
Journal of Zhejiang University:Agriculture and Life Sciences
基金
国家自然科学基金(31402352
3140131018)
国家星火计划(2015GA701031
2014GA701031)
浙江省宁波市自然科学基金(2015610131
2014A610185)
浙江省自然科学基金(LY17C190008)
浙江省教育厅课题(Y201432753)
农业农村部设施农业装备与信息化重点实验室开放课题
关键词
图像处理
压缩感知
鱼群行为
低溶氧胁迫
异常行为检测
image processing
compressed sensing
fish school behavior
low dissolved oxygen stress
abnormal behavior detection
作者简介
卢焕达(https://orcid.org/0000-0002-4924-1038),E-mail:huandalu@163.com。;通信作者(Corresponding author):于欣(https://orcid.org/0000-0002-8648-0812),E-mail:yuxin@zju.edu.cn。
