摘要
茶叶嫩芽分选鉴定是实现茶叶自动分级的重要环节,为实现对优质茶叶嫩芽的准确识别,提出了一种基于改进目标检测算法YOLOv7的茶叶嫩芽分级识别方法。首先,在主干网络融入CBAM注意力机制,构建CBAM-ELAN模块,在通道和空间注意模块的协同作用下,降低背景部分的权重,增强对嫩芽特征的提取能力;然后,在Neck侧引入感受野增强(RFE)模块和显式视觉中心(EVC)模块,生成新的感受野关系,提升层内特征的调节能力,增强对嫩芽特征的提取能力。改进YOLOv7算法对于单芽等级茶叶的检测精度均值为89.3%,一芽一叶检测精度均值为88.9%,一芽二叶检测精度均值为95.7%,与原始YOLOv7算法相比,检测精度均值分别提高了1.3、1.3和1.5个百分点。该方法在兼顾准确率的前提下,实现了端到端的目标检测和对优质茶苗不同姿态的识别,可为茶叶嫩芽的分级和识别提供理论基础。
Tea bud classification and identification is an important part of the realization of automatic tea classifier.In order to achieve accurate identification of high-quality tea shoots,a tea bud classification method based on improved YOLOv7 was proposed.Firstly,CBAM-ELAN module is constructed based on the CBAM attention mechanism integrated into the backbone network.Under the synergistic effect of channel and spatial attention module,the weight of background part is reduced and the feature extraction ability of bud is enhanced.Receptive field enhancement(RFE)module and explicit visual center(EVC)module on the Neck side are introduced to generate a new receptive field relationship.The ability to adjust the features in the layer was improved and the ability to extract bud features was enhanced.In this paper,the detection accuracy of the improved YOLOv7 algorithm for tea leaves is 89.3%for single bud,88.9%for one bud and one leaf,and 95.7%for one bud and two leaves.Compared with the original YOLOv7,the accuracy rate is increased by 1.3,1.3 and 1.5 percentage points respectively.End-to-end target detection and different posture recognition of high-quality tea seedlings were realized,which could provide an important theoretical basis for tea bud classification and recognition.
作者
李龙
孙雅
LI Long;SUN Ya(Application and Demonstration Base of Innovative Methods,Anhui University of Science and Technology,Huainan 232001,China;School of Artificial Intelligence,Anhui University of Science and Technology,Huainan 232001,China)
出处
《河南工程学院学报（自然科学版）》
2024年第3期67-73,共7页
Journal of Henan University of Engineering：Natural Science Edition
基金
安徽省创新方法推广应用与示范基地开放基金(2022AHIMG03)
安徽省高校自然科学研究计划重点项目(KJ2021A0418)
安徽理工大学高层次人才引进科研启动基金(13200391)。
