摘要
为了鱼苗的饲养、运输和销售过程中对一定数量或批量的鱼苗实现精确计数,提出了一种基于机器视觉跟踪的鱼苗计数算法。按一定时间间隔动态跟踪拍摄鱼苗图像,采用图像处理技术提取每个图像块的面积作为其特征值;根据面积的大小,将图像块划分为正常图像和异常图像,并对正常图像和异常图像分别进行分析处理,进而建立图像特征值与鱼苗的数量之间的相关算法模型;利用动态跟踪拍摄的多组图像数据回归获得最优模型参数,得到估计函数,并应用到计数中累加求和。实验结果表明:基于动态视觉跟踪和回归分析的鱼苗计数方法,能够有效区分鱼苗计数过程中的黏连问题,提高了鱼苗计数精度。
Aiming at accurate counting of a number or batch of fish in process of feeding,transporting and selling,a fish counting algorithm based on machine vision tracking is proposed. According to a certain time interval,dynamic track and shoot fish images,use image processing technology to extract area of each image block as its eigenvalue. According to the size of area,the image block is divided into normal image and abnormal image,and the normal image and abnormal image are respectively analyzed and processed. Correlated algorithm model between the image eigenvalue and the number of fish is established. The optimal model parameters are obtained by regression of multiple sets of image data captured by dynamic tracking,determine the estimation function and applied to the count summation. Experimental results show that this method of fry counting based on dynamic visual tracking and regression analysis can effectively distinguish the adhesion problems during fry counting and improve precision of fry counting.
出处
《传感器与微系统》
CSCD
2018年第2期154-157,160,共5页
Transducer and Microsystem Technologies
基金
宁波市自然科学基金资助项目(2016A610210)
宁波市科技攻关项目(2015C50061)
宁波市公益重大专项项目(2015C110015)
宁波市产业技术应用重大专项项目(2016B10017)
关键词
精确计数
视觉跟踪
图像面积
算法模型
accurate counting
visual tracking
image area
algorithm model
