基于改进YOLOv5_OBB的中华绒螯蟹旋转目标检测  

Rotating target detection of Chinese Eriocheir sinernsis based on improved YOLOv5_OBB

作　　者：袁红春[1] 白宝来 陶磊 YUAN Hongchun;BAI Baolai;TAO Lei(School of Information,Shanghai Ocean University,Shanghai 201306,China)

机构地区：[1]上海海洋大学信息学院,上海201306

出　　处：《福建农林大学学报（自然科学版）》2024年第2期284-288,共5页Journal of Fujian Agriculture and Forestry University:Natural Science Edition

基　　金：国家自然科学基金项目(41776142)。

摘　　要：【目的】提出一种基于改进YOLOv5_OBB的旋转目标检测方法,快速、准确地检测和定位中华绒螯蟹。【方法】首先,在YOLOv5_OBB的主干网络中引入高效通道注意模块;其次,采用BiFPN网络结构进行特征融合模块设计,实现高效的双向跨尺度连接和加权特征融合;最后,采用变焦损失(varifocal loss)解决正负样本不均衡问题。【结果】改进后YOLOv5_OBB模型的P(precision)、R(recall)和mAP(mean average precision)分别达到95.4%、95.2%和90.1%,比原模型分别提高了1.0%、1.9%和1.3%。【结论】该模型能够实时、准确地检测和定位中华绒螯蟹,实现自动化养殖。【Objective】An improved YOLOv5_OBB-based rotating object detection method was proposed for rapid and accurate detection and localization of Chinese Eriocheir sinernsis.【Method】Firstly,an efficient channel attention module was introduced into the backbone network of YOLOv5_OBB to enhance the model's detection performance for small targets.Secondly,Bi-directional feature pyramid network(BiFPN)structure was employed in the feature fusion module to achieve efficient bi-directional cross-scale connections and weighted feature fusion.Finally,varifocal loss was utilized to address the issue of imbalance between positive and negative samples,thereby improving the model's detection accuracy for dense targets.【Result】Experimental results demonstrated that the improved YOLOv5_OBB model achieved precision(P),recall(R),and mean average precision(mAP)of 95.4%,95.2%,and 90.1%,respectively.The model can effectively and accurately detect and locate Chinese E.sinernsis in real time.【Conclusion】The model can accurately detect and locate Chinese E.sinernsis in real time,complete the primary task of automated breeding.

关 键 词：中华绒螯蟹 YOLOv5_OBB 旋转目标检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] S966.16[自动化与计算机技术—计算机科学与技术]