School of Mechanical and Intelligent Mamufacturing,Central South University of Forestry and Technology,Changsha 410004,China
Show less
文章历史+
Received
Published
2025-12-15
2026-09-20
Issue Date
2026-11-13
PDF (7954K)
摘要
针对油茶树枝条自动化修剪的视觉感知问题,尤其在室外自然环境中,枝条易受风扰、枝叶遮挡等环境因素影响,设计开发一种基于改进YOLOv8n的视觉伺服修剪平台。平台分为视觉检测系统和机械臂修剪系统。在视觉检测系统中,分别引入高效上卷积模块(efficient up-convolution block,EUCB)与多层感知器(multilayer perceptron,MLP)模块,设计EM-YOLO(EUCB and MLP-you only look once)模型,对特征融合模块与上采样模块进行改进,提升模型在复杂情况下对形态多变目标枝的识别能力;设计一种基于预测框底部中心点的定位策略,运用图像二值化与骨架提取技术来预测修剪姿态;基于标准D-H法建立机械臂运动学模型;通过图像误差建立视觉伺服控制策略,对目标枝实现动态逼近与姿态自校正。试验结果与原YOLOv8n相比,EM-YOLO对油茶目标枝的精确率、召回率、mAP0.5分别提高0.5%、7.1%、3.5%,研究采集的不同光照下目标枝样本检测性能保持稳定。室内自动化修剪试验中,平台实现了对油茶树目标枝的检测与修剪,识别成功率为81.29%,漏检率为14.03%,在1~2 m作业范围内空间定位误差小于1 cm,平台总体修剪成功率为80.9%,能够有效实现对油茶树目标枝自动化剪枝作业,为油茶树的智能化、规范化管理提供技术参考。
Abstract
Aiming at the visual perception problem of automatic pruning of Camellia oleifera branches, especially in the outdoor natural environment, branches are vulnerable to wind disturbance, branches and leaves shading and other environmental factors, a visual servo pruning platform based on improved YOLOv8n is designed and developed. The platform is divided into visual inspection system and manipulator pruning system. In the visual inspection system, efficient up-convolution block (EUCB) and multilayer perceptron (MLP) modules are introduced respectively, and EM-YOLO (EUCB and MLP-you only look once) model is designed to improve the feature fusion module and up sampling module, so as to enhance the recognition ability of the model in complex situations. A positioning strategy based on the center point at the bottom of the prediction frame is designed, and the image binarization and skeleton extraction technology are used to predict the pruning posture. The kinematics model of the manipulator is established based on the standard D-H method. The visual servo control strategy is established through the image error to achieve dynamic approximation and attitude self-tuning for the target branch. Compared with the original YOLOv8n, the accuracy, recall and mAP0.5 of EM-YOLO for Camellia oleifera target branches are improved by 0.5%, 7.1% and 3.5%, respectively. The detection performance of EM-YOLO in target branch samples collected under different lighting conditions remain stable. In the indoor automatic pruning experiment, the platform has realized the detection and pruning of the target branches of Camellia oleifera, the recognition success rate is 81.29%, the missing detection rate is 14.03%, the spatial positioning error within the operation range of 1-2 m is less than 1 cm, and the overall pruning success rate of the platform is 80.9%, which can effectively realize the automatic pruning of the target branches of Camellia oleifera, and provide technical reference for the intelligent and standardized management of Camellia oleifera.
YOLO(you only look once)作为单阶段目标检测算法的代表,在准确率与实时性上具有一定的优势。原始YOLOv8采用端到端的回归算法实现整体图像检测,在图像特征提取过程中,主要利用固定大小的卷积核来进行扫描,对检测对象形态变化特征表达能力较差。自然环境下生长的油茶树枝、叶、果三者互相重叠,遮挡问题严重,油茶生长密集且树枝形状多样,原始YOLOv8无法较好地实现对油茶树枝、叶及果的3种重叠目标的识别。为提高油茶树徒长枝的检测精度,在YOLOv8n的基础上加以改进,通过引入MLP模块和EUCB模块,从特征提取和解码2个阶段同时加以改进和提高,提高检测精度和实时性。
为进一步提升EM-YOLO模型的整体性能,针对二值化网络中信息丢失和梯度错误的问题,本研究引入二值化MLP模块。该模块通过移位操作方法解决信息丢失问题并改善特征表示能力[25],从而增强模型在复杂场景下对目标枝的识别与检测效果。二值化MLP模块的结构见图6(a),包含2个二值化线性的主层,并通过移位操作来优化特征图的表示,其中a1和a2表示不同步长的移位,例如移动一个像素或多个像素,用于生成多尺度空间特征,见图6(b)。该模块通过拟归一化(BN)和修正线性单元(rescaled linear unit,RPReLU)激活函数处理输入的特征以减少信息的丢失和梯度消失。图6(c)和6(d)分别为水平移位和垂直移位、拼接移位操作的具体实现。
对于自动化修剪作业,末端修剪刀具的姿态控制是完成精准修剪的关键[28-29]。根据农艺要求,修剪刀具的目标点通常在枝条基部或靠近主干处,以获得较为平整的切口,利于树体恢复,需要求解末端修剪刀具的最佳切入角度,即对目标枝条的姿态进行估计。本研究设计一种轻量化的位姿估计策略,由带目标枝的检测框图像得到目标枝的感兴趣区域(region of interest,ROI)并进行二值化与骨架提取处理,见图8。利用最小二乘法拟合出代表树枝主轴方向的直线方程。计算直线与图像水平轴的夹角作为偏航角估计,其计算公式为
在运动学模型验证可靠的基础上,进一步建立基于视觉反馈的伺服控制方法,将图像特征误差实时反馈到关节空间运动[33-34]。将视觉模块输出的修剪点空间坐标()、枝条主轴偏航角()、通过相机标定获得的相机内参矩阵( K )以及手眼变换矩阵作为输入端,得到实时的末端期望位姿。在控制过程中,将修剪点在平面图像上的像素误差与姿态偏差共同构成伺服控制误差信号()。建立图像上的图像雅可比矩阵,控制器通过图像雅可比矩阵建立线性控制律。公式为
GaoL, JinL, LiuQ,et al.Recent advances in the extraction,composition analysis and bioactivity of Camellia (Camellia oleifera Abel.) oil[J].Trends in Food Science &; Technology,2024,143:104211.
[2]
张金云,吴世军,刘姚,等.果树整形修剪技术[J].安徽农学通报,2024,30(3):39-43.
[3]
ZhangJinyun, WuShijun, LiuYao,et al.Fruit tree shaping and pruning technology[J].Anhui Agricultural Bulletin,2024,30(3):39-43.
[4]
解子安.油茶树体精准调控技术研究[D].长沙:中南林业科技大学,2025.
[5]
XieZian.Study on precise regulation technology of Camellia oleifera tree[D].Changsha:Central South University of Forestry and Technology,2025.
LiuJia, YangLiling, MaWenqiang,et al.Current situation and trend of fruit tree pruning machinery[J].Journal of Agricultural Mechanization Research,2025,47(1):262-268.
GuWenjuan, LiuHaozhuang, WeiJin,et al.Tomato fruit recognition in complex scenes based on FPBW-YOLO v8[J].Transactions of the Chinese Society for Agricultural Machinery,2025,56(8):467-478.
LiaoJuan, LiuKaixuan, YangYuqing,et al.Rice disease recognition in natural environment based on RDN-YOLO[J].Transactions of the Chinese Society for Agricultural Machinery,2024,55(8):233-242.
LiuMengshu, ZhangChunqi, ChaoJinyang,et al.Improved wheat disease detection system based on YOLO v8n[J].Transactions of the Chinese Society for Agricultural Machinery,2024,55(S1):280-287,355.
WangBinbin, YangGuijun, YangHao,et al.UAV images for detecting maize tassel based on YOLO_X and transfer learning[J].Transactions of the Chinese Society of Agricultural Engineering,2022,38(15):53-62.
WangDeng, HeLeiying, DuXiaoqiang,et al.Rapid recognition of winter bamboo shoots based on improved YOLOv8n and ground penetrating radar images[J].Journal of Agricultural Mechanization Research,2026,48(5):151-159.
PengXuan, ZhouJianping, XuYan,et al.Cotton top bud recognition method based on YOLOv5-CPP in complex environment[J].Transactions of the Chinese Society of Agricultural Engineering,2023,39(16):191-197.
YuanChao, MaJianing, ZhangPanhao,et al.Design and experiment of the visual servo planting platform forphalaenopsis tissue-cultured seedlings using improved YOLOv8[J].Transactions of the Chinese Society of Agricultural Engineering,2024,40(20):138-146.
YueKai, ZhangPengchao, WangLei,et al.Recognizing citrus in complex environment using improved YOLOv8n[J].Transactions of the Chinese Society of Agricultural Engineering,2024,40(8):152-158.
GeYiyuan, LiAo, MengQingxiang,et al.Small target fruit recognition and localization method for soft jujube kiwifruit based on YOLOv8-MI[J].Journal of Agricultural Mechanization Research,2026,48(4):249-257.
[26]
马保建.矮化休眠期枣树智能剪枝关键技术研究[D].杭州:浙江大学,2022.
[27]
MaBaojian.Study on the key technologies of intelligent pruning for dwarf and dormant jujube tree[D].Hangzhou:Zhejiang University,2022.
HuangZhe, TangShixi, ShenGuandong,et al.Depth recognition of branch obstacles of apple picking robot based on improved YOLOv4[J].Hubei agricultural science,2024,63(8):10-16,22.
ZhouHongwei, JiHaowen, WuYixuan,et al.Improve the detection model of tree decay by dioryctria based on the YOLOv8[J].Forest Engineering,2025,41(1):126-137.
[32]
LinG, ZhuL, LiJ,et al.Collision⁃free path planning for a guava⁃harvesting robot based on recurrent deep reinforcement learning[J].Computers and Electronics in Agriculture,2021,188:106350.
[33]
WangY, LiuD, ZhaoH,et al.Rapid citrus harvesting motion planning with pre⁃harvesting point and quad⁃tree[J].Computers and Electronics in Agriculture,2022,202:107348.
[34]
ChenB, GongL, YuC,et al.Workspace decomposition based path planning for fruit⁃picking robot in complex greenhouse environment[J].Computers and Electronics in Agriculture,2023,215:108353.
ZhangQin, LiuFengpu, JiangXianping,et al.Motion planning method and experiments of tomato bunch harvesting manipulator[J].Transactions of the Chinese Society of Agricultural Engineering,2021,37(9):149-156.
WangR, ChenL, ChenJ,et al. Different nitrate and ammonium ratios affect growth and physiological characteristics of camellia oleifera abel.seedlings[J].Forests,2018,9(12):784.
[40]
董娅兰.葡萄冬季剪枝点识别与定位方法研究[D].乌鲁木齐:新疆大学,2023.
[41]
DongYalan.Research on identification and localization methods of winter grapevine pruning point[D].Urumqi:Xinjiang University,2023.
[42]
RahmanM M, MunirM, MarculescuR.EMCAD:efficient multi-scale convolutional attention decoding for medical image segmentation[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),June 16-22,2024,Seattle,WA,USA,IEEE,2024.
[43]
GaoT, ZhangZ, ZhangY,et al.BHViT:binarized hybrid vision transformer[C]//2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),June 10-17,2025,Nashville,TN,USA,IEEE.2025.
FanXiangpeng, ZhangYanqi, ZhouShuo,et al.Recognition and position method of tomato picking robot based on improved YOLOv8s and RGB-D information fusion[J].Transactions of the Chinese Society of Agricultural Engineering,2025,41(15):106-116.
[46]
ZhangZ Y.A flexible new technique for camera calibration[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2000,22(11):1330-1334.
[47]
牟毅.丘陵山区茶园修剪机器人结构设计与关键技术研究[D].贵阳:贵州大学,2025.
[48]
MouYi.Research on structure design and key technology of tea garden pruning robot in Hilly and mountainous areas[D].Guiyang:Guizhou University,2025.
[49]
崔书彬.电动式五自由度绿篱修剪机械臂的设计与研究[D].呼和浩特:内蒙古农业大学,2023.
[50]
CuiShubin.Design and research of electrodynamic five-degree-of-freedom robotic arms for hedge trimming[D].Hohhot:Inner Mongolia Agricultural University,2023.
[51]
周志华.基于深度学习的辣椒识别及采摘技术研究[D].成都:电子科技大学,2025.
[52]
ZhouZhihua.Research on chili pepper identification and harvesting technology based on deep learning[D].Chengdu:University of Electronic Science and technology,2025.
[53]
AbbasT, AbbarA, AbboodM.Kinematics analysis of 5 DOF robotic arm[J].Engineering and Technology Journal,2020,38(3A):412-422.
[54]
翁天浩.林区智能作业车臂架仿真分析及优化设计[D].长沙:中南林业科技大学,2024.
[55]
WengTianhao.Simulation analysis and optimization design of the arm frame of intelligent work vehicles in forest areas[D].Changsha:Central South University of forestry and technology,2024.
RenXiaolin.Research on image-based visual servoing positioning and tracking control technology of manipulator[D].Changchun:University of Chinese Academy of Sciences (Changchun Institute of optics,precision mechanics and physics,Chinese Academy of Sciences),2021.
[58]
李斌.基于视觉伺服系统的机械臂定位控制方法研究[D].成都:电子科技大学,2025.
[59]
LiBin.Research on manipulator positioning control method based on visual servo system[D].Chengdu:University of Electronic Science and technology,2025.
[60]
TsuchidaS, LuH, KamiyaT,et al.Characteristics based visual servo for 6DOF robot arm control[J].Cognitive Robotics,2021,1:76-82.