[1]施武,袁伟皓,杨梦道,等.一种基于改进YOLOv8n-seg的轻量化茶树嫩芽的茶梗识别模型[J].江苏农业学报,2025,(01):75-86.[doi:doi:10.3969/j.issn.1000-4440.2025.01.010]
　SHI Wu,YUAN Weihao,YANG Mengdao,et al.A lightweight model for identifying the stalks of tea buds based on the improved YOLOv8n-seg[J].,2025,(01):75-86.[doi:doi:10.3969/j.issn.1000-4440.2025.01.010]
点击复制

一种基于改进YOLOv8n-seg的轻量化茶树嫩芽的茶梗识别模型()
分享到：

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年01期
页码:
75-86
栏目:
农业信息工程
出版日期:
2025-01-31

文章信息/Info

Title:
A lightweight model for identifying the stalks of tea buds based on the improved YOLOv8n-seg
作者:
施武; 袁伟皓; 杨梦道; 许高建
(安徽农业大学信息与人工智能学院,安徽合肥230036)
Author(s):
SHI Wu; YUAN Weihao; YANG Mengdao; XU Gaojian
(School of Information and Artificial Intelligence, Anhui Agricultural University, Hefei 230036, China)
关键词:
图像识别; 茶叶采摘; 轻量化模型; YOLOv8n-seg; VanillaNet
Keywords:
image recognition; tea harvesting; lightweight model; YOLOv8n-seg; VanillaNet
分类号:
TP212；S571.1
DOI:
doi:10.3969/j.issn.1000-4440.2025.01.010
文献标志码:
A
摘要:
茶树嫩芽茶梗识别对实现茶叶采摘的自动化和智能化具有重要意义。然而,现有的目标检测算法检测茶树嫩芽茶梗存在精度较低、计算量大、模型体积庞大等问题,限制了其在终端设备上的部署。因此,本研究基于YOLOv8n-seg模型,提出一种轻量化的茶树嫩芽茶梗识别模型YOLOv8n-seg-VLS,并在以下3个方面进行了改进：引入VanillaNet轻量化模块替代原有卷积层,以降低模型的复杂程度；在颈部引入大型可分离核注意力模块(LSKA),以降低存储量和计算资源消耗；将YOLOv8的损失函数从中心点与边界框的重叠联合(CIoU)替换为边界框自身形状与自身尺度之间的损失(Shape-IoU),从而提高边界框的定位精度。在采集的茶叶数据集上进行测试,结果表明,改进后获得的YOLOv8n-seg-VLS模型的平均精度值(mAP)方面表现较好,交并比阈值为0.50的平均精度值（mAP0.50）为94.02%,交并比阈值为0.50至0.95的平均精度值（mAP0.50∶0.95）为62.34%；模型的准确度(P)为90.08%,召回率(R)为89.96%；改进模型的每秒传输帧数(FPS)为245.20帧,模型的大小为3.92 MB,仅为YOLOv8n-seg大小的57.39%。研究结果为后续茶叶智能化采摘装备的研发提供了技术支持。
Abstract:
Identifying the stalks of tea buds is of great significance for achieving automated and intelligent tea picking. However, existing object detection algorithms face significant challenges in terms of low detection accuracy, high computational demands, and large model sizes, which collectively limit their deployment on edge devices. To address these challenges, we proposed a lightweight tea stalk detection model, YOLOv8n-seg-VLS, which was based on the YOLOv8n-seg framework. The model incorporated three significant enhancements. First, the VanillaNet lightweight module was introduced to replace traditional convolutional layers, thereby reducing the model’s complexity. Second, a large separable kernel attention (LSKA) module was incorporated into the neck section of the network to minimize memory usage and resource consumption. Third, the loss function of YOLOv8 was modified from center intersection over union (CIoU) to shape- and scale-aware intersection over union (Shape-IoU), thereby enhancing the precision of bounding box localization. The experimental results on a collected tea dataset demonstrated that YOLOv8n-seg-VLS achieved a mean average precision (mAP) of 94.02% at mAP0.50 and 62.34% at mAP0.50∶0.95, with a precision of 90.08% and a recall of 89.96%. In comparison to the original YOLOv8n-seg, the proposed model demonstrated an improvement in frame rate, reaching 245.20 frames per second (FPS). Moreover, the model size was 3.92 MB, which was only 57.39% of the size of YOLOv8n-seg. These results provide technical support for further development of intelligent tea harvesting equipment.

参考文献/References:

[1]李杨,董春旺,陈建能,等. 茶叶智能采摘技术研究进展与展望[J]. 中国茶叶,2022,44(7):1-9.
[2]金寿祥,周宏平,姜洪喆,等. 采摘机器人视觉系统研究进展[J]. 江苏农业学报,2023,39(2):582-595.
[3]裴伟,王晓林. 基于图像信息的茶叶二维采摘坐标的提取[J]. 浙江农业学报,2016,28(3):522-527.
[4]邵佩迪,吴明晖,季亚波,等. 茶叶嫩芽机器视觉识别算法研究[J]. 农业装备与车辆工程,2020,58(4):34-36,45.
[5]张博. 基于RGB-D的茶叶识别与定位技术研究[D]. 沈阳：沈阳工业大学,2020.
[6]罗坤,吴正敏,曹成茂,等. 茶鲜叶嫩梢捏切组合式采摘器设计与试验[J]. 农业工程学报,2022,38(13):1-9.
[7]龙樟,姜倩,王健,等. 茶叶嫩芽视觉识别与采摘点定位方法研究[J]. 传感器与微系统,2022,41(2):39-41,45.
[8]李翰林,高延峰,熊根良,等. 基于实例分割的大场景下茶叶嫩芽轮廓提取与采摘点定位[J]. 农业工程学报,2024,40(15):135-142.
[9]吕军,方梦瑞,姚青,等. 基于区域亮度自适应校正的茶 叶嫩芽检测模型[J]. 农业工程学报,2021,37(22):278-285.
[10]杨大勇,黄正栎,郑昌贤,等. 基于改进YOLOv8n的茶叶嫩稍检测方法[J]. 农业工程学报,2024,40(12):165-173，313.
[11]SHUAI L Y, MU J, JIANG X Q, et al. An improved YOLOv5-based method for multi-species tea shoot detection and picking point location in complex backgrounds[J]. Biosystems Engineering,2023(231):117-132.
[12]WANG T, ZHANG K M, ZHANG W, et al. Tea picking point detection and location based on Mask-RCNN[J]. Information Processing in Agriculture,2023,10(2):267-275.
[13]YAN L J, WU K H, LIN J, et al. Identification and picking point positioning of tender tea shoots based on MR3P-TS model[J]. Frontiers in Plant Science,2022(13):962391.
[14]XU F, LI B, XU S. Accurate and rapid localization of tea bud leaf picking point based on YOLOv8[C]//MENG X F, CHEN Y, SUO L M, et al. 2023 China National Conference on Big Data and Social Computing. Urumqi, China：Springer,2023:261-274.
[15]高洋. 基于深度学习的茶叶采摘点定位[D]. 合肥：安徽农业大学,2023.
[16]RUSSELL B C, TORRALBA A, MURPHY K P. LabelMe:a database and web-based tool for image annotation[J]. International Journal of Computer Vision,2008,77(1):157-173.
[17]WANG Y, LI B, YUAN X L. BrightFormer:a transformer to brighten the image[J]. Computers & Graphics,2023(110):49-57.
[18]XIAO T, XIA T, YANG Y, et al. Learning from massive noisy labeled data for image classification[C]//HORST B, DAVID F, CORDELIA S, et al. 2015 IEEE conference on computer vision and pattern recognition. Boston,MA,USA：IEEE,2015:2691-2699.
[19]ZHONG Z, ZHENG L, KANG G L, et al. Random erasing data augmentation[J]. Proceedings of the AAAI Conference on Artificial Intelligence,2020,34(7):13001-13008.
[20]YUE X, QI K, NA X Y, et al. Improved YOLOv8-seg network for instance segmentation of healthy and diseased tomato plants in the growth stage[J]. Agriculture,2023,13(8):1643.
[21] LI H W, HUANG J Z, GU Z N, et al. Positioning of mango picking point using an improved YOLOv8 architecture with object detection and instance segmentation[J]. Biosystems Engineering,2024(247):202-220.
[22]GE Z, LIU S T, WANG F, et al. YOLOX:exceeding YOLO series in 2021[EB/OL]. (2021-07-18)
[2024-08-20]. https://doi.org/10.48550/arXiv.2107.08430.
[23]CHEN H T, WANG Y H, GUO J Y, et al. VanillaNet:the power of minimalism in deep learning[EB/OL]. (2023-05-23)
[2024-08-20]. https://doi.org/10.48550/arXiv.2305.12972.
[24]LAU K W, PO L M, REHMAN Y A. Large separable kernel attention:rethinking the large kernel attention design in CNN[J]. Expert Systems with Applications,2024(236):121352.
[25]ZHENG Z H, WANG P, REN D W, et al. Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics,2022,52(8):8574-8586.
[26]ZHANG H, ZHANG S J. Shape-IoU:more accurate metric considering bounding box shape and scale[EB/OL]. (2023-12-29)
[2024-08-20]. https://doi.org/10.48550/arXiv.2312.17663.
[27] LU J B, YU M M, LIU J Y. Lightweight strip steel defect detection algorithm based on improved YOLOv7[J]. Scientific Reports,2024,14(1):13267.
[28]SANDLER M, HOWARD A, ZHU M L, et al. MobileNetV2:inverted residuals and linear bottlenecks[C]//BROWN M S, MORSE B, PELEG S, et al. 2018 IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City,USA：IEEE,2018:4510-4520.
[29] PAOLETTI M E, HAUT J M, PEREIRA N S, et al. GhostNet for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing,2021,59(12):10378-10393.
[30]MA N, ZHANG X, ZHENG H T, et al. ShuffleNetV2:practical guidelines for efficient CNN architecture design[C]//LIN T Y, PATTERSON G, RONCHI M R, et al. 2018 European Conference on Computer Vision. Munich,Germany：Springer,2018:116-131.
[31]WANG D D, HE D J. Fusion of mask R-CNN and attention mechanism for instance segmentation of apples under complex background[J]. Computers and Electronics in Agriculture,2022,196:106864.
[32]WANG M J, LI Y, MENG H W, et al. Small target tea bud detection based on improved YOLOv5 in complex background[J]. Frontiers in Plant Science,2024,15:1393138.
[33]WANG C Y, BOCHKOVSKIY A, LIAO H Y. YOLOv7:trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//BROWN M S, LI F F, MORI G, et al. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver,BC,Canada：IEEE,2023:7464-7475.
[34]YE R, SHAO G Q, GAO Q, et al. CR-YOLOv9:improved YOLOv9 multi-stage strawberry fruit maturity detection application integrated with CRNET[J]. Foods,2024,13(16):2571.

相似文献/References:

[1]孙云云,江朝晖,董伟,等.基于卷积神经网络和小样本的茶树病害图像识别[J].江苏农业学报,2019,(01):48.[doi:doi:10.3969/j.issn.1000-4440.2019.01.007]
　SUN Yun-yun,JIANG Zhao-hui,DONG Wei,et al.Image recognition of tea plant disease based on convolutional neural network and small samples[J].,2019,(01):48.[doi:doi:10.3969/j.issn.1000-4440.2019.01.007]
[2]邱洪涛,孙裴,侯金波,等.基于Caffe的猪肉新鲜度分级的设计与实现[J].江苏农业学报,2019,(02):461.[doi:doi:10.3969/j.issn.1000-4440.2019.02.029]
　QIU Hong-tao,SUN Pei,HOU Jin-bo,et al.Design and implementation of pork freshness grading based on Caffe[J].,2019,(01):461.[doi:doi:10.3969/j.issn.1000-4440.2019.02.029]
[3]牛学德,高丙朋,南新元,等.基于改进DenseNet卷积神经网络的番茄叶片病害检测[J].江苏农业学报,2022,38(01):129.[doi:doi:10.3969/j.issn.1000-4440.2022.01.015]
　NIU Xue-de,GAO Bing-peng,NAN Xin-yuan,et al.Detection of tomato leaf disease based on improved DenseNet convolutional neural network[J].,2022,38(01):129.[doi:doi:10.3969/j.issn.1000-4440.2022.01.015]
[4]梁凯博,孙立,汪禹治,等.基于超轻量化卷积神经网络的番茄病虫害诊断[J].江苏农业学报,2024,(03):438.[doi:doi:10.3969/j.issn.1000-4440.2024.03.006]
　LIANG Kai-bo,SUN Li,WANG Yu-zhi,et al.Diagnosis of tomato pests and diseases based on super lightweight convolutional neural network[J].,2024,(01):438.[doi:doi:10.3969/j.issn.1000-4440.2024.03.006]
[5]蒋东山,刘金洋,张浩淼,等.基于CNN和Transformer的绿豆干旱胁迫识别模型[J].江苏农业学报,2025,(01):87.[doi:doi:10.3969/j.issn.1000-4440.2025.01.011]
　JIANG Dongshan,LIU Jinyang,ZHANG Haomiao,et al.Drought stress recognition model of mung bean based on CNN and Transformer[J].,2025,(01):87.[doi:doi:10.3969/j.issn.1000-4440.2025.01.011]

备注/Memo

备注/Memo:
收稿日期：2024-08-23基金项目：安徽省高校自然科学研究重点项目(KJ2020A0106)；安徽省重大科技专项(202103b06020013)；安徽省大学生创新创业计划项目(S202310364126)作者简介：施武(2004-),男,安徽六安人,本科,研究方向为计算机视觉。(Tel)15656038325；(E-mail)wshi@stu.ahau.edu.cn通讯作者：许高建,(E-mail)xugj@ahau.edu.cn
更新日期/Last Update: 2025-02-28