摘要
本文旨在研究基于深度网络模型的茶叶嫩芽识别方法,根据茶叶的品级和质量要求,把茶叶嫩芽分为一芽一叶和一芽两叶,因为茶叶的生长姿态千差万别,所以又在茶叶嫩芽识别模型中加入关于遮挡情况的分类。选用了基于VGG-16,ResNet-50和ResNet-101特征提取网络的Faster R-CNN深度网络模型分别对茶叶嫩芽数据样本进行训练,同时,该方法与三种相同特征提取网络的SSD深度网络模型进行对比,实验结果表明,基于VGG-16特征提取网络的Faster R-CNN深度网络模型的识别效果较好﹐得出茶叶嫩芽识别模型的精确度为85.14%,召回率为78.90,mAP为82.17%。该深度网络模型能够有效识别茶叶嫩芽目标,为茶叶的智能化采摘提供了技术支撑。
At present,domestic tea picking methods are mainly manual picking,supplemented by mechanical picking.Manual picking has low efficiency and high economic cost.Although mechanical picking is efficient,it will damage the buds.The artificial intelligence that has emerged in recent years and applied to tea picking can effectively solve the problem of mechanical picking and damaged buds,one of the most important tasks is to realize automatic and efficient identification of tea buds.The purpose of this paper is to study the method of tea bud recognition based on the depth network model.According to the grade and quality requirements of tea,the tea buds are divided into one bud,one leaf and one bud,two leaves.Because the growth posture of tea is very different,the classification of occlusion is added to the tea bud recognition model.The Faster R-CNN deep network model based on the VGG-16,ResNet-50 and ResNet-101 feature extraction network is selected to train the tea bud data samples respectively.At the same time,the method is compared with the SSD deep network model of the three same feature extraction networks.The experimental results show that the Faster R-CNN deep network model based on the VGG-16 feature extraction network has a good recognition effect.The accuracy of tea bud recognition model is 85.14%,recall rate is 78.90,and map is 82.17%.The depth network model can effectively identify the target of tea buds and provide technical support for intelligent tea picking.
作者
许高建
张蕴
赖小燚
XU Gao-jian;ZHANG Yun;LAI Xiao-yi(School of information and computer,Anhui Agricultural University,Hefei 230036,China;Guochuang rail Technology Co.,Ltd,Zhuzhou 412000,China)
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2020年第11期1131-1139,共9页
Journal of Optoelectronics·Laser
基金
农业农村部农业部引进国际先进农业科学技术948项目(2015-Z44,2016-X34)
安徽省高校自然科学基金资助项目。
