安徽科技学院学报,2023,37(3):96-101JournalofAnhuiScienceandTechnologyUniversity收稿日期:2022-12-07基金项目:安徽省重点研究与开发项目(202104f06020019)。作者简介:张琪(1987-),女,安徽濉溪人,硕士研究生,主要从事智能农业装备研究。通信作者:曹浩,教授,E-mail:3655505689@163.com。基于改进的FasterR-CNN苹果缺陷区域目标检测张琪1,曹浩2*(1.安徽科技学院机械工程学院,安徽凤阳233100;2.安徽科技学院信息与网络工程学院,安徽凤阳233100)摘要:目的:为实现在复杂果园环境下,初摘苹果快速、准确的分拣,基于FasterR-CNN模型,提出一种识别苹果损伤区域的算法模型。方法:修改FasterR-CNN模型特征提取网络,以苹果损坏区域为研究对象,将缺陷区域从图像中快速检测出来。结果:应用改进模型,对苹果缺陷区域检测的平均精度值较高(87.06%)。结论:应用改进的FasterR-CNN算法可以实现复杂环境下苹果较小缺陷区域快速检测,为果园苹果快速粗分拣设备开发提供技术支撑。关键词:FasterR-CNN;苹果缺陷区域;目标检测;图像处理中图分类号:TP249文献标志码:A文章编号:1673-8772(2023)03-0096-06开放科学(资源服务)标识码(OSID):DOI:10.19608/j.cnki.1673-8772.2023.0051ImprovedFasterR-CNNbasedonappledefectiveregiontargetdetectionZHANGQi1,CAOHao2*(1.CollegeofMechanicalEngineering,AnhuiScienceandTechnologyUniversity,Fengyang233100,China;2.CollegeofInformation&NetworkEngineering,AnhuiScienceandTechnologyUniversity,Fengyang233100,China)Abstract:Objective:Toachievefastandaccuratesortingoffirst-pickedapplesinacomplexorcharden-vironment,analgorithmicmodelforidentifyingappledamagedregionsisproposedbasedontheFasterR-CNNmodel.Methods:TheFasterR-CNNmodelfeatureextractionnetworkismodifiedtostudytheappledamagedareatodetectthedefectiveareafromtheimagequickly.Results:Theapplicationoftheimprovedmodelresultesinahighaverageaccuracyvalueof87.06%forappledefectregiondetection.Conclusion:TheapplicationoftheimprovedFasterR-CNNalgorithmcanachieverapiddetectionofsmallerdefectareasofapplesincomplexenvironmentsandprovidetechnicalsupportforthedevelopmentofrapidcoarsesortingequipmentforapplesinorchards.Keywords:FasterR-CNN;Appledefectregion;Targetdetection;Imageprocessing我国是农产品生产大国,苹果产...
