摘要
为了在田间环境下准确检测食用玫瑰花及其成熟度,实现花期玫瑰花的自动化采摘,针对田间光照、遮挡等因素造成识别精度较差的问题,提出了一种基于YOLOv5s的改进模型,对花蕾期、采摘期、败花期食用玫瑰花的生长状态进行检测。首先,为了增强多尺度特征融合能力,对特征融合结构进行改进。其次,采用多分支结构训练提高精度,在颈部网络C3模块进行改进。最后,为了提升特征信息的提取能力,在颈部网络中添加融合注意力模块,使模型关注检测目标,减少玫瑰花的误检及漏检现象。改进后的模型检测总体类别平均精度较原始模型提升了3.6个百分点,达到90.4%,对3个花期玫瑰花的检测精度均有提升。本研究结果为非结构环境下的不同花期食用玫瑰花检测提供了更加准确的方法。
In order to accurately detect edible roses and their maturity in the field and realize the automatic picking of flowering roses,an improved model based on YOLOv5s was proposed to solve the problem of poor recognition accuracy caused by factors such as light and occlusion in the field.The growth state of edible roses at bud,picking and abortive flowering stages was detected.Firstly,in order to enhance the ability of multi-scale feature fusion,the feature fusion structure was improved.Secondly,multi-branch structure training was used to improve the accuracy,and the neck network C3 module was improved.Finally,in order to improve the ability of feature information extraction,a fusion attention module was added to the neck network to make the model focus on the detection target and reduce the false detection and missed detection of roses.The mean average precision of the improved model was 3.6 percentage points higher than that of the original model,reaching 90.4%,and the detection accuracy of roses in three flowering periods was improved.The results of this study provide a more accurate method for detecting edible roses at different flowering stages in unstructured environment.
作者
化春键
黄宇峰
蒋毅
俞建峰
陈莹
HUA Chunjian;HUANG Yufeng;JIANG Yi;YU Jianfeng;CHEN Ying(School of Mechanical Engineering,Jiangnan University,Wuxi 214122,China;Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment&Technology,Wuxi 214122,China;School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)
出处
《江苏农业学报》
CSCD
北大核心
2024年第8期1464-1472,共9页
Jiangsu Journal of Agricultural Sciences
基金
国家自然科学基金项目(62173160)。
