基于改进TV-L1模型的虾苗计数方法  

A Method on Shrimp Counting Based on Improved TV-L1 Model

作　　者：季玉瑶 魏伟波[1] 赵增芳[2] 杨振宇 JI Yu-yao;WEI Wei-bo;ZHAO Zeng-fang;YANG Zhen-yu(College of Computer Science and Technology,Qingdao University,Qingdao 266071,China;Institute of Oceanology Chinese Academy of Sciences(IOCAS),Qingdao 266071,China)

机构地区：[1]青岛大学计算机科学与技术学院,青岛266071 [2]中国科学院海洋研究所,青岛266071

出　　处：《青岛大学学报（自然科学版）》2018年第4期62-68,82,共8页Journal of Qingdao University(Natural Science Edition)

基　　金：国家自然科学基金(批准号:61772294)资助

摘　　要：为实现渔业生产中对虾苗的计数需求,针对虾苗图像光照不均、图像粘连等问题,对图像复原的TV-L1模型进行了改进,提出了一种新的对数商图像连通区域标记面积计数法。利用改进TV-L1模型作为低通滤波算子对图像进行平滑滤波和图像光照分量估计,将在对数域定义的原图像与光照分量估计的商作为光照归一化图像,对该图像进行形态学处理,采用最大类间方差法获取虾苗分割的最佳阈值并进行二值化处理,最后利用连通区域标记面积法来确定虾苗数量。实验结果表明,该方法的计数平均准确率较高。In order to realize the counting demand of shrimp in fishery production,the TV-L1 model of image restoration is improved and a new connected component labeling-counting algorithm in marking area is proposed to solve the problems of uneven illumination and image adhesion.Improved TV-L1 model is used as a low-pass filter to smooth images,which estimates the illumination component.In log files,the quotient of the original image is defined and its illumination component estimation as illumination normalization image,and make mathematical morphology for this image.The optimal threshold segmentation of shrimp is obtained by OTSU method as well as binaryzation processing.The number of shrimp area is determined by the connected component labeling-counting algorithm.The experiment shows that the average accuracy rate of paper which put forward method is higher.

关 键 词：虾苗计数 改进的TV-L1模型 光照归一化 形态学处理 最大类间方差法 连通区域标记面积法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]