王 硕1,2，范良忠2，刘 鹰3
（1太原科技大学电子信息工程学院，山西 太原030024；
2浙江大学宁波理工学院，浙江 宁波 315100；
3中国科学院海洋研究所，山东 青岛 266071）

摘要：对单帧大菱鲆鱼苗计数时，由于鱼苗的重叠，导致图像中的鱼苗粘连。为此采用计算机视觉的方法，提出了一种基于曲线演化的图像处理方法来解决鱼苗图像粘连问题，并完成对鱼苗的准确计数。将图像预处理后的二值化鱼苗粘连图像的轮廓曲线作为初始演化曲线，通过水平集方法对曲线进行常值演化，统计演化过程中最大轮廓曲线数作为鱼苗数量。对选取的10帧粘连比较严重的二值化鱼苗图像分别采用面积法、连通域法、曲线演化法进行分离计数研究。结果表明：曲线演化法在鱼体数目增多时仍有很好的分离计数效果，有利于使用数鱼机对鱼苗进行准确计数。

关键词：大菱鲆鱼苗；鱼苗计数；计算机视觉；曲线演化
 

The research of turbot fry counting method based on computer vision
WANG Shuo1,2, FAN Liangzhong2, LIU Ying3
(1 School of Electronic Information Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China;
2 Ningbo Institute of Technology, Zhejiang University, Ningbo 315100, China;
3 Institute of Oceanology, China Academy of Sciences, Qingdao266071,China)

Abstract: For single frame turbot fish count, the overlapping of the turbot fish produce the images of the adhesion phenomenon. Adopting the method of computer vision, this paper proposes a image processing method based on curve evolution to solve the single frame turbot fish image of adhesion and complete accurate count of the fry, which will be carried out after the image pre-processing, the contour curve of binary adhesion fish as the initial evolution curve, and through the level set method constant evolution is carried out, then counting the largest number of contour curve can be the number of fry. To select the 10 frames of severe adhesion binary fish image, the methods of area, connected domain separation counting method and the curve evolution method were applied to study the separation and counting. The experimental results showed that the curve evolution method had a good effect on separation and counting even when the fry number increases, which is beneficial for using fish counting machine to count fry accurately.

Keywords: turbot fiy; fish count; computer vision; curve evolution

（来源：《渔业现代化》2015,42(1):16-19）