摘要
为提高斑玉蕈(Hypsizygus marmoreus)生产过程中质量监测能力,基于深度学习和图像识别技术构建菌盖斑点表型识别分析系统。采用MV-CA013-20GC工业相机采集斑玉蕈菌盖图像,并通过图像颜色校正(标准色块提取、白平衡校正和Lab色域转换与色差优化)消除环境光干扰,确保图像色彩还原准确,采用X-AnyLabeling 2.5.3对500张图像进行边界框标注,构建训练数据集,基于YOLOv5检测模型开展模型训练,训练完成后选用树莓派4B部署平台实现菌盖自动分割与菌盖斑点特征分析,采用OpenCV的findContours算法提取斑点数量、大小、分布状态和颜色。该系统可有效完成菌盖斑点表型识别与分析。
In order to improve quality monitoring during the production process of Hypsizygus marmoreus,an image recognition and analysis system for phenotypic characteristics of spots on pilei was developed based on deep learning.The system was constructed through image acquisition,color calibration,bounding box annotation,model training,model deployment,phenotypic criteria establishment,and image-based analytical modules.First,an MV-CA013-20GC industrial camera was used to capture pileus images.Color correction algorithms(white balance and Lab color difference optimization)were applied to eliminate interference of ambient light and ensure color fidelity.Then,500 color-corrected images were annotated by the X-AnyLabeling tool to construct a training dataset.A YOLOv5 object detection model was trained with data augmentation techniques to improve generalization performance.The trained model was then deployed on a Raspberry Pie 4B computing device to enable automated pileus segmentation and spot feature analysis.Phenotypic parameters of pileus spots,including spot count,size,spatial distribution,and color attributes,were quantified using the findContours algorithm in OpenCV.This system provided a technical framework for standardized phenotypic analysis of pileus spots of H.marmoreus fruiting bodies.
作者
姜宁
陈静
余昌霞
苏海辉
董浩然
李巧珍
于海龙
李正鹏
JIANG Ning;CHEN Jing;YU Changxia;SU Haihui;DONG Haoran;LI Qiaozhen;YU Hailong;LI Zhengpeng(Institute of Edible Fungi,Shanghai Academy of Agricultural Sciences,National Research Center Biotechnology and Engineering of Edible Fungi,Shanghai 201403,China;Shanghai Finc Biotech Incorporated Company,Shanghai 201401,China;Zhangzhou Lehui Intelligent Technology Limited Company,Zhangzhou 363599,Fujian,China)
出处
《食用菌学报》
北大核心
2025年第4期81-90,共10页
Acta Edulis Fungi
基金
国家现代农业产业技术体系(CARS-20)。
