基于计算机视觉的鱼类形态轮廓特征自动提取OA北大核心CSTPCD
Automatic extraction of contour features of fish morphology based on computer vision
鱼类形态变化多样,其形态轮廓特征具有种的特异性,并作为鱼类识别和分类的重要科学依据.形态轮廓特征的提取效果直接影响到自动识别鱼类的精确度,因此,为了研究计算机视觉对鱼类形态轮廓特征的自动提取效果,根据 2017 年 9-11 月在太平洋海域采集的1 尾大眼金枪鱼的二维图像,进行计算机视觉分析.通过对鱼类图像进行灰度转换,双边滤波,二值化图像处理和轮廓提取等图像处理.利用 8 个方位的链码技术对鱼类轮廓进行链码信息的自动提取.通过椭圆傅里叶变换计算出形态信息系数,并对鱼类形态进行轮廓重建.结果显示,金枪鱼图像处理后能较好地得到轮廓图像,其链码信息会随着鱼类形态轮廓像素的大小发生变化,而鱼类形态的轮廓重建随着谐次的变化而变化.研究表明,自动提取鱼类形态轮廓特征效果较好.鱼类形态系数在低谐次变化波动较大,在高谐次变化波动较小.轮廓重建在低谐次变换对鱼类整体轮廓信息影响较大,在高谐次变换对鱼类局部轮廓信息影响较大.研究结果为鱼类自动识别和分类奠定前期基础,也为其他相关自动化研究提供借鉴和参考.
With the rapid development of artificial intelligence,modern fish biology research technology has been constantly updated.Automated and intelligent fish identification will help promote modern fish biology research development.The contour feature of fish morphology is one of the important features of fish recognition.Fish morphology is diversed,and the contour features of fish morphology have species specificity.Meanwhile,it serves as an important scientific basis for fish identification and classification.The extraction effects of morphological contour features directly affect the accuracy of automatic fish identification.Therefore,in order to study the effect of computer vision on the automatic extraction of fish morphological contour features,a two-dimensional image of one tail T.obesus was collected in the Pacific Ocean from September to November 2017 for computer vision ana-lysis through the fish image gray level transformation,bilateral filter,binary image processing and contour extrac-tion and other image processing.8-direction chain code technology was used to automatically extract the chain code information of fish contour.The morphological information coefficient was calculated by elliptic Fourier transform and the contour reconstruction was carried out.The results revealed that the contour image of tuna could be obtained well after processing,and the chain code information changes with the size of the contour pixel of fish shape,and the contour reconstruction of fish shape changed with the change of harmonic number.The research results showed that the automatic extraction of fish contour features was effective.The morphological coefficients of fish fluctuated greatly in low harmonic number and little in high harmonic number.The low harmonic number transformation had a great influence on the overall contour information of fish,while the high harmonic number transformation had a great influence on the local contour information of fish.The results of this study lay a prelim-inary foundation for automatic fish recognition and classification,and also provide references for other related automation research.
欧利国;蓝振峰;刘必林;陈新军;陈勇
上海海洋大学海洋生物资源与管理学院,上海 201306上海海洋大学信息学院,上海 201306上海海洋大学海洋生物资源与管理学院,上海 201306||上海海洋大学,大洋渔业资源可持续开发教育部重点实验室,上海 201306||上海海洋大学,国家远洋渔业工程技术研究中心,上海 201306||上海海洋大学,农业农村部大洋渔业开发重点实验室,上海 201306
水产学
鱼类计算机视觉形态轮廓链码形态系数轮廓重建自动提取
fishescomputer visionmorphological contourchain codemorphological coefficientcontour recon-structionautomatic extraction
《水产学报》 2024 (012)
61-71 / 11
国家重点研发计划(2019YFD0901404);国家自然科学基金(41876141);上海市高校特聘教授"东方学者"岗位计划(0810000243);农业农村部西北太平洋公海渔业资源综合科学调查专项(D-8021-21-0109-01);上海市科技创新行动计划(19DZ1207502) National Key R&D Program of China(2019YFDO901404);National Natural Science Found-ation of China(41876141);Program for Professor of Special Appointment(Eastern Scholar)at Shanghai Institu-tions of Higher Learning under Contract(0810000243);Ministry of Agriculture and Rural Affairs Major Program of Fisheries Resource Scientific Survey in the Northwest Pacific Ocean(D-8021-21-0109-01);Shanghai Science and Technology Innovation Action Plan(19DZ1207502)
评论