一种基于YOLO v5的克氏原螯虾性别检测方法  

A Gender Detection Method for Crayfish Based on YOLO v5

作　　者：孔得溦 李尚 陈义明 KONG De-wei;LI Shang;CHEN Yi-ming(College of Information and Intelligence,Hunan Agricultural University,Changsha 410128,PRC)

机构地区：[1]湖南农业大学信息与智能科学技术学院,湖南长沙410128

出　　处：《湖南农业科学》2024年第3期59-63,共5页Hunan Agricultural Sciences

基　　金：国家现代农业产业技术体系专项(CARS-48);湖南省重点研发计划项目(2020NK2033)。

摘　　要：针对传统人工判别克氏原螯虾性别效率低、成本高的问题,提出了一种基于YOLO v5的克氏原螯虾性别检测模型,实现了克氏原螯虾性别特征的自动判别。采用自主设计装置拍摄克氏原螯虾图像,使用Labelme工具进行基于雄虾交接器检测和基于区域特征检测两种方法的数据标注,在Pytorch框架下以Resnet-18为特征提取网络训练二分类模型,基于YOLO v5训练交接器检测和区域特征检测两种模型。结果表明:基于区域特征检测的模型具有较高的检测性能和准确性,能够高效、低成本地提取克氏原螯虾性别特征,对克氏原螯虾品种改良具有重要意义。In response to the low efficiency and high cost of traditional manual discrimination of the gender of crayfish(Procambarus clarkii),this study proposed a gender detection model for crayfish based on YOLO v5,achieving automatic discrimination of the gender of crayfish.Crayfish images were captured using a self-designed device,and data annotation was performed using two methods,i.e.,detection based on the petasma and detection based on regional features,using the Labelme tool.A binary classification model was trained using Resnet-18 as the feature extraction network under the Pytorch framework,and two models,one for petasma detection and the other for regional feature detection,were trained based on YOLO v5.The results showed that the model based on regional feature detection had high detection performance and accuracy,efficiently and cost-effectively extracting the gender characteristics of crayfish,which was of great significance for the breeding improvement of crayfish.

关 键 词：YOLO v5 目标检测 克氏原螯虾 深度学习 卷积神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]