基于改进YOLOv8n的番茄果实检测方法  

Tomato maturity detection method based on improved YOLOv8n

作　　者：赵承敏 余宏杰 孙豪杰 ZHAO Chengmin;YU Hongjie;SUN Haojie(College of Intelligent Manufacturing,Anhui Science and Technology University,Fengyang 233100,China;College of Information and Network Engineering,Anhui Science and Technology University,Bengbu 233000,China)

机构地区：[1]安徽科技学院智能制造学院,安徽凤阳233100 [2]安徽科技学院信息与网络工程学院,安徽蚌埠233000

出　　处：《安徽科技学院学报》2025年第2期59-66,共8页Journal of Anhui Science and Technology University

基　　金：安徽省重点研究与开发计划(202204c06020065)。

摘　　要：针对不同情况下番茄果实识别检测精度低和存在漏检的问题,提出一种基于改进YOLOv8n的番茄检测模型,旨在提高番茄的识别准确率和检测效率。首先,在特征提取过程中集成了CBAM(Convolutional Block Attention Module)注意力机制模块。其次,引入BIFPN-SID模块进一步提高模型的泛化能力。再结合损失函数Inner-SIoU优化模型的训练过程。试验表明,所提出的模型在番茄果实识别任务的性能指标显示,准确率高达94.4%,召回率为91.3%,F 1值为92.8%,平均精度均值达到95.6%。与原模型YOLOv8n相比分别提高5.8、6.4、6.1、3.3个百分点。在检测准确率与速度方面,改进后的模型均表现出色,可以应用于番茄采摘机器人目标识别检测。In view of the low accuracy and missed detection of tomato fruit under different conditions,a tomato detection model based on improved YOLOv8n was proposed,which aimed to improve the identification accuracy and detection efficiency of tomato.Firstly,the CBAM(Convolutional Block Attention Module)attention mechanism is incorporated during feature extraction.Additionally,the BIFPN-SID module was introduced to enhance the model's generalization capability.Then,the Inner-SIoU loss function was employed to optimize the model's training process.Experiments showed that the accuracy of the proposed model in the tomato fruit recognition task reached 94.4%,while the recall rate was 91.3%.Meanwhile,the F 1 value was 92.8%,and the average accuracy amounted to 95.6%,respectively.Compared to the baseline YOLOv8n model,there had been a respective increase of 5.8%,6.4%,6.1%and 3.3%,respectively.In terms of detection accuracy and speed,the improved model had excellent performance and could be applied to target recognition detection of tomato picking robots.

关 键 词：YOLOv8n 目标检测 BIFPN-SID 注意力机制 Inner-SIoU 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]