基于YOLO-ODM的温室草莓成熟度的快速检测  

YOLO-ODM based rapid detection of strawberry ripeness in greenhouse

作　　者：陈仁凡 谢知[1,2] 林晨 CHEN Renfan;XIE Zhi;LIN Chen(College of Mechanical and Electronic Engineering,Fujian Agriculture and Forestry University,Fuzhou 350002,China;Fujian Province Key Laboratory of Agricultural Information Perception Technology,Fuzhou 350002,China)

机构地区：[1]福建农林大学机电工程学院,福州350002 [2]福建省农业信息感知技术重点实验室,福州350002

出　　处：《华中农业大学学报》2023年第4期262-269,共8页Journal of Huazhong Agricultural University

基　　金：福建省自然科学基金项目(2019J01403)。

摘　　要：为解决温室环境下草莓果实快速准确识别问题,提出一种基于改进YOLOv5s的草莓成熟度检测方法。在主干中引入Shuffle_Block作为特征提取网络,从而实现模型轻量化。同时,在颈部结构中使用全维度动态卷积模块(omni-dimensional dynamic convolution,ODConv),以提高网络对草莓目标的信息挖掘能力,降低计算量,并进一步实现轻量化。结果显示,改进后的YOLO-ODM(YOLO with ODConv module)模型的平均精度均值达97.4%,模型体积是7.79 Mb,在GPU上的单张平均检测时间仅11 ms,浮点运算量为6.9×10^(9)。与原网络相比,轻量化的YOLO-ODM方法在提高检测精度的同时,模型大小缩减43%,浮点运算量降低52%。以上结果表明,该轻量化方法可快速准确地对温室环境下草莓果实的成熟度进行检测,实现草莓的生长状态监测。An improved YOLOv5s-based method for rapidly detecting strawberry ripeness was proposed to solve the problem of rapid and accurate identification of strawberry fruits in greenhouse.The Shuffle_Block was introduced as a feature extraction network in the backbone to lightweight the model.Meanwhile,the omni-dimensional dynamic convolution(ODConv)module was used in the neck structure to enhance the information mining ability of model for strawberry targets,reduce computational complexity,and further achieve lightweight.The results showed that the average precision of the improved YOLO-ODM model reached 97.4%.The model size is 7.79 Mb.The average detection time on the GPU is only 11 ms per image,and the floating-point operations are 6.9×10^(9) FLOPs.Compared with the original network,the lightweighted YOLO-ODM method improved the accuracy of detection while reducing model size by 43% and floating-point operations by 52%.It is indicated that the lightweighted method can rapidly and accurately detect the ripeness of strawberry fruit in greenhouse,monitor the growth status of strawberries.

关 键 词：草莓 成熟度 目标检测 YOLOv5s Shuffle_Block 精准识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]