基于改进YOLOv8n一Pose的鱼苗体长测量研究

打开文本图片集

中图分类号：S969；TP391 文献标识码：A 文章编号：2095-5553（2025）12-0128-08

Abstract：Accuratemeasurementoffish frybody length iscritical forthedevelopmentof smartaquaculture.Traditional manual methods are time-consuming，labor-intensive，and may harm fish fry. This study used 20-100 mm grass carp fry （Ctenopharyngodon idella） as test subjects.An image acquisition platform was developed to obtain RGB - D data from the top view of the fry.Weproposed ROS—YOLO，which replaces the original C2f module of YOLOv8n—Pose with reparameterized convolution-based shufle one-shot aggregation（RCS—OSA）and introduces a simple atention module （SimAM）intothe main feature extractionlayer，todetect key bodylength pointsoffish fry.Depth information for 3D keypointcordinatetransformation wasobtained through thedepthmap.Finally，fish frybodylengthswerecalculated based on keypoint coordinates.In experiments，ROS—YOLO achievedanaverage keypoint detection accuracy of 99.2% ，with 3.97M parameters and 125 frame/s. Compared to manual measurements，the overall average error in automatic measurement results was 2.87mm and the average absolute error was 5.85% ：

Keywords：grasscarp fry；frybody length；atention mechanism；three-dimensional coordinates；smart aquaculture

0 引言

状况，而且还可以作为饲料投喂、分级养殖、捕捞售卖以及鱼类生物量估算的重要依据，对鱼苗养殖期间的生产效率评估和智慧管理具有重要意义。（剩余13117字）