摘要
金枪鱼属(Thunnus)鱼类是我国远洋渔业极为重要的渔获资源,其表型纹理信息不仅具有鱼种的特异性,而且可作为分类的科学依据。传统鱼类纹理特征分析主要是定性描述分析,而计算机视觉技术可为鱼类纹理特征提供定量分析数据。因此,本研究通过计算机视觉对3种金枪鱼图像进行预先定位基准点,通过移动基准点确定纹理特征区域并自动截取。对纹理图像进行灰度转换和灰度量化处理,量化的灰度图像进行灰度共生矩阵计算,并对灰度共生矩阵进行归一化处理。通过归一化的灰度共生矩阵计算出6个纹理指标,并分析纹理指标的距离和方向的变化趋势,通过因子分析研究金枪鱼纹理指标。研究结果表明,通过计算机视觉的纹理分析,3种金枪鱼纹理指标提取效果较好,其纹理指标在距离值为4时,变化趋势趋于稳定,而3种金枪鱼的纹理指标方向变化,其均值方向具有代表性。3种金枪鱼的因子分析,第1主成分贡献率为81.10%,表明提取的6个纹理指标意义较大且效果较好。以期为金枪鱼智能识别奠定前期基础,也为其他鱼类表型纹理研究提供借鉴和参考。
Tuna is an important commercial product and accounts for a large proportion of the world’s fisheries.At the same time,tuna also has a great impact on the development of China’s fishery production.The genus Thunnus is an important catch resource in China’s pelagic fishery,and its phenotype texture information is not only characteristic of fish species,but also can be used as a scientific basis for classification.Traditional fish texture feature analysis has mainly used qualitative description analysis(QDA),but computer vision technology can provide quantitative analysis(QA)data instead.This paper used computer vision to pre-locate the basic standard point of the images of three Thunnus species,determining the texture feature regions by moving the basic standard point and automatically acquiring them.The texture image was transformed into and quantized with gray level.The quantized gray level image was used for gray level co-occurrence matrix(GLCM)calculation,and the obtained GLCM was normalized.Six texture indexes were calculated by normalized gray co-occurrence matrix,and the variation trend of the distance and direction of texture indexes was analyzed.The texture indexes of the genus Thunnus were studied by factor analysis(FA).Through texture analysis of computer vision,the results show that the extraction effect of texture index for the three Thunnus species was good.When the distance value was 4,the change trend of texture index tended to be stable,the direction of texture index of the three Thunnus species changed,and its mean direction was representative.Factor analysis of the three Thunnus species shows that the contribution rate of the first principal component was 81.10%,indicating that the extracted six texture indexes had high significance and good effect.This work lays a preliminary foundation for the intelligent recognition of the genus Thunnus,and also provides a reference for other fish phenotype texture research.
作者
欧利国
李汶龙
刘必林
陈新军
陈勇
石一茜
侯庆联
OU Liguo;LI Wenlong;LIU Bilin;CHEN Xinjun;CHEN Yong;SHI Yixi;HOU Qinglian(College of Marine Sciences,Shanghai Ocean University,Shanghai 201306,China;College of Information Technology,Shanghai Ocean University,Shanghai 201306,China;Key Laboratory of Sustainable Exploitation of Oceanic Fisheries Resources,Ministry of Education,Shanghai 201306,China;National Distant-water Fisheries Engineering Research Center,Shanghai Ocean University,Shanghai 201306,China;Key Laboratory of Oceanic Fisheries Exploration,Ministry of Agriculture and Rural Affairs,Shanghai 201306,China)
出处
《中国水产科学》
CAS
CSCD
北大核心
2022年第5期770-780,共11页
Journal of Fishery Sciences of China
基金
国家重点研发计划项目(2019YFD0901404)
国家自然科学基金项目(41876141)
上海市高校特聘教授“东方学者”岗位计划项目(0810000243)
农业农村部全球渔业资源调查监测评估项目(D-8021-21-0109-01)
上海市科技创新行动计划项目(19DZ1207502)
关键词
计算机视觉
金枪鱼属
纹理特征
灰度共生矩阵
纹理指标
因子分析
computer vision
Thunnus
texture features
gray level co-occurrence matrix
texture index
factor analysis
作者简介
欧利国(1992-),男,博士研究生,从事渔业资源生物学与智慧渔业学研究.E-mail:919989412@qq.com;通信作者:刘必林,教授,从事渔业资源生物学与智慧渔业学研究.E-mail:bl-liu@shou.edu.cn
