Abstract: The measurement of fish caudal peduncle in aquaculture farm is of great significance in aquaculture management. At present, the main measuring method is depend on manual work which is inefficient and subjective. This study proposed a new measuring method based on image processing and linear fitting. Firstly, the images are collected in the fish farm using equipments; then, the external contour of the fish is obtained through edge detection and contour extraction based on the fish morphological character; finally, the length of caudal peduncle is figured out using corner detection and least square line fitting after the extracting of ROI of the caudal peduncle. The results show that the average measurement time is 1.01 seconds, the average measurement error is 0.34mm, and the accuracy of measurement has reached over 95%. Lots of studies have shown that the proposed method can not only solve the problem that manual operation is complex as well as subjective, but also avoid time-consuming measurement; and provide an effective way for non-contact mass measurement of fish caudal peduncle.

Key words: fish caudal peduncle, length measurement, image processing, least square line fitting, Liusha port of Zhanjiang

摘要： 水产养殖场内的鱼体尾柄测量在养殖管理中有着重要意义。目前的尾柄测量方法主要依靠人工，操作复杂、效率低，且结果主观性强。研究提出一种基于图像处理和线性拟合的鱼体尾柄测量方法。首先在养殖场通过装置采集图像，通过边缘检测和轮廓提取得到鱼体的外部轮廓，基于鱼的形态特征这一先验知识，提取到尾柄感兴趣区域，最后对尾柄区域进行角点检测和最小二乘直线拟合，得到尾柄线长度。结果显示，该方法平均测量耗时1.01s，平均测量偏差0.34 mm，测量精度达到95%以上。采用该方法不仅可以解决人工测量操作繁琐、主观性强等问题，而且能够避免测量过程耗时过长，为鱼体尾柄的大批量非接触式测量提供有效新途径。

关键词: 鱼体尾柄, 长度测量, 图像处理, 最小二乘直线拟合, 湛江流沙港