基于AD-YOLOX-Nano的茶叶嫩芽识别算法  

Tea bud recognition algorithm based on AD-YOLOX-Nano

作　　者：高芳征 温鑫 黄家才 陈光明[3] 金少宇 赵雪迪 Gao Fangzheng;Wen Xin;Huang Jiacai;Chen Guangming;Jin Shaoyu;Zhao Xuedi(School of Automation,Nanjing Institute of Technology,Nanjing,211167,China;School of Mechanical Engineering,Nanjing Institute of Technology,Nanjing,211167,China;College of Engineering,Nanjing Agricultural University,Nanjing,210000,China)

机构地区：[1]南京工程学院自动化学院,南京市211167 [2]南京工程学院机械工程学院,南京市211167 [3]南京农业大学工学院,南京市210000

出　　处：《中国农机化学报》2025年第1期178-184,F0002,共8页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金面上项目(61873120);江苏省重点研发项目课题(BE2021016—5);江苏省自然科学基金面上项目(BK20201469);江苏省高等学校自然科学研究重大项目(20KJA510007);江苏省研究生实践创新计划项目(SJCX22_1061)。

摘　　要：为解决茶叶嫩芽识别困难,提高自然环境下茶叶嫩芽识别的精确性和鲁棒性,提出一种融入注意力机制和深度可分离卷积的改进型YOLOX-Nano(AD-YOLOX-Nano)茶叶嫩芽识别算法。该算法以YOLOX-Nano模型为基础,采用CSPDarkNet作为主干网络,通过在CSPDarkNet网络中引入深度可分离卷积(Depthwise Separable Convolution)来减少特征提取工作量,并将卷积注意力模块(Convolutional Block Attention Module)融入到YOLOX-Nano网络的特征金字塔中,学习不同通道的特征相关性,增强网络的深度信息传递,提高模型在不同场景下对茶叶嫩芽的识别能力。结果表明:AD-YOLOX-Nano算法的平均精度AP值和F_(1)值分别为85.6%和86%,相较于同环境下YOLOX-Nano算法,该算法的模型大小基本保持不变,但其AP值和F_(1)值分别提高2.7%和3%。与常用的YOLOv5-S、YOLOv4和Faster R-CNN等目标检测算法相比,该AD-YOLOX-Nano算法模型大小仅为它们的1/7,但AP值分别提高5.4%、5.5%和6.28%。所提算法在模型轻量化和检测精度方面优势显著,为茶叶智能化采摘的嵌入式硬件部署提供有效解决方案。In order to address the difficulties in identifying tea buds and improve the accuracy and robustness of tea bud recognition in natural environments,this paper develops an improved YOLOX-Nano algorithm(AD-YOLOX-Nano)by integrating attention mechanism and depthwise separable convolution.The algorithm is based on the YOLOX-Nano model,which uses CSPDarkNet as the backbone network.It reduces the workload of feature extraction by introducing Depthwise Separable Convolution in the CSPDarkNet network.The Convolutional Block Attention Module is incorporated into the feature pyramid of the YOLOX-Nano network to learn the feature correlation of different channels,enhance the transmission of depth information in the network,and improve the recognition capability of tea buds in different scenarios.The experimental results show that the AP value and F_(1) values of the AD-YOLOX-Nano algorithm are 85.6%and 86%,respectively.Compared to the YOLOX-Nano algorithm in the same environment,this algorithm maintains a similar model size,but achieves an improvement of 2.7%in AP value and 3%in F_(1) value.Compared to the commonly used object detection algorithms such as YOLOv5-S,YOLOv4,and Faster R-CNN,the AD-YOLOX-Nano algorithm has a model size of only 1/7 of theirs.But it achieves an improvement of 5.4%,5.5%,and 6.28%in AP value,respectively.The proposed algorithm has significant advantages in terms of model lightweighting and detection accuracy,providing an effective solution for the deployment of embedded hardware for intelligent tea picking.

关 键 词：茶叶嫩芽识别 AD-YOLOX-Nano算法 注意力机制 深度可分离卷积 

分 类 号：S24[农业科学—农业电气化与自动化] TP391.4[农业科学—农业工程]