摘要
目的:在设施农业种植环境下,提出一种基于改进YOLOv5模型的目标检测方法,实现草莓常见的几种病害及养分不足情况的快速准确检测。方法:通过将C3模块替换为C2f模块、增加SimAM注意力机制、增加检测层等对原始YOLOv5模型结构进行改进,以草莓植株叶片、花、果为研究对象,应用改进YOLOv5模型提取草莓植株生长特征,并进行病害及养分状况快速检测。结果:应用改进YOLOv5模型对草莓植株生长状况进行识别的准确率较高,各识别项的平均检测精度均值(mAP@0.5)为97.9%,较原始YOLOv5和YOLOv8模型分别高出13.6%和10.3%,识别图像传输帧率较原YOLOv5模型提升了22 f/s。结论:该模型可以实现设施农业环境下草莓病害和养分状况的识别与检测,具有较高识别精度和较快识别速度,适用于移动底盘端部署,为草莓智能化精准喷雾作业提供一定的技术支撑。
Objective:To achieve rapid and accurate detection of several common strawberry diseases and nutrient deficiencies in a facility agricultural planting environment,a target detection method based on an improved YOLOv5 model was proposed.Methods:By replacing the C2f module with the C3 module,adding the SimAM attention mechanism,and adding detection layers,the original YOLOv5 model structure was improved.The leaves,flowers,and fruits of strawberry plants were studied,and the improved YOLOv5 model was applied to extract growth characteristics of strawberry plants and quickly detect diseases and nutrient conditions.Results:The improved YOLOv5 model had a high accuracy in identifying the growth status of strawberry plants,and the average detection accuracy of each recognition item averaged(mAP@0.5)97.9%,which was 13.6%and 10.3%higher than the original YOLOv5 and YOLOv8 models,respectively.The recognition image transmission frame rate was improved by 22 f/s compared to the original YOLOv5 model.Conclusion:This model could realize the identification and detection of strawberry diseases and nutrient status under the facility agricultural environment,with high recognition accuracy and fast recognition speed,and was suitable for mobile chassis deployment,providing a certain technical support for intelligent and accurate spray of strawberries.
作者
刘苏杭
张华
朱婷倩
吴凡
童以
LIU Suhang;ZHANG Hua;ZHU Tingqian;WU Fan;TONG Yi(College of Mechanical Engineering,Anhui Science and Technology University,Fengyang 233100,China)
出处
《安徽科技学院学报》
2024年第2期87-94,共8页
Journal of Anhui Science and Technology University
基金
安徽省科技特派员农业物质技术装备领域揭榜挂帅项目(2022296906020001)
安徽科技学院功能农业专项(2021gnny01)。
