基于YOLOv5s的番茄果实检测模型改进  

Improvement of Tomato Fruit Detection Model Based on YOLOv5s

作　　者：赵真 ZHAO Zhen(School of Automation,Qingdao University,Qingdao 266000,China)

机构地区：[1]青岛大学自动化学院,青岛266000

出　　处：《自动化与仪表》2025年第4期109-112,117,共5页Automation & Instrumentation

摘　　要：为了提升采摘机器人采摘番茄果实的识别准确率,减少误判和漏检的情况,该文提出一种基于YOLOv5s的番茄检测改进模型,首先将CA注意力机制加入到YOLOv5s模型的特征提取环节,为了进一步提高模型的泛化能力,实现多尺度特征融合,引入了BiFPN结构替换原有的PANet结构,同时结合损失函数IoU进一步优化模型的训练过程,从而提高检测精度。经过试验验证,提出的模型在番茄果实检测任务中获得了显著成果。相比YOLOv5s模型,改进后的模型准确率、召回率、F1分数和平均精度分别上升2.2%、2.1%、2.2%和1.3%。改进后的模型可以应用于番茄采摘机器人,并可以对成熟番茄与未成熟番茄进行分类,与原YOLOv5s算法相比,有效提升了检测的精度。In order to improve the recognition accuracy of tomato fruit picking robots and reduce misjudgments and missed detections,this paper proposes an improved tomato detection model based on YOLOv5s.Firstly,the CA attention mechanism is added to the feature extraction stage of the YOLOv5s model.In order to further improve the generalization ability of the model and achieve multi-scale feature fusion,the BiFPN structure is introduced to replace the original PANet structure,and the training process of the model is further optimized by combining the loss function IoU,thereby improving the detection accuracy.After experimental verification,the proposed model has achieved significant results in tomato fruit detection tasks.Compared to the YOLOv5s model,the improved model has increased accuracy,recall,F1 score,and average precision by 2.2%,2.1%,2.2%and 1.3%,respectively.The improved model can be applied to tomato harvesting robots and can classify mature and immature tomatoes.Compared with the original YOLOv5s algorithm,it effectively improves the detection accuracy.

关 键 词：YOLOv5s 注意力机制 采摘机器人 BiFPN 目标检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]