Accurate identification of wooden chair legs and localization of grasping positions are crucial for achieving efficient robotic loading operations. However, existing instance segmentation-based image processing algorithms often struggle to balance accuracy and speed, and their large number of parameters hinders practical deployment in industrial scenarios. Furthermore, these algorithms are prone to missed detections and false detections in complex environments. To address the aforementioned issues, this paper proposes an improved YOLOv11n-seg model. By incorporating lightweight modules and a novel loss function, the model achieves a favorable balance between accuracy and efficiency, thereby optimizing its deployment performance. The specific improvements are as follows: firstly, RepViT is introduced as the backbone network to enhance feature extraction capability while reducing computational complexity. Secondly, the lightweight dynamic upsampler DySample is adopted in the neck network to improve feature fusion efficiency. Finally, Shape-IoU is used to replace the traditional CIoU Loss loss function, optimizing the regression accuracy of bounding boxes. Results of ablation experiments and comparative experiments demonstrate that the improved YOLOv11n-seg algorithm outperforms the original model in both inference speed and segmentation accuracy. Specifically, the total number of model parameters is reduced by 11.31%, the floating-point operations (FLOPs) are decreased by 9.62%, and the mean average precision for segmentation (mAP@0.5:0.95) is increased by 4.61 percentage points. These findings validate the effectiveness of the improved algorithm and provide technical support for the subsequent precise positioning and grasping of wooden chair legs by robots.
LiYayun.Design of a combined production line for CNC loading and unloading based on industrial robots[J].Mechanical Engineering & Automation,2025,54(2):156-158.
[3]
HintonG E, SalakhutdinovR R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[4]
GowrammaB H, PoornimaB, JadekarR B.Enhancing silkworm breeding through advanced image segmentation:a comparative study of U-net,mask R-CNN,and FCN[C]//2024 First International Conference on Innovations in Communications,Electrical and Computer Engineering,(ICICEC). October 24-25,2024 ,Davangere,India.IEEE,2024:1-6.
[5]
ChenL-C, PapandreouG, KokkinosI,et al.DeepLab:semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2018,40(4):834-848.
[6]
YuQ S, XuX R, LiuY Z,et al.Robot plane grasping pose detection based on U2-net[C]//2023 IEEE International Conference on Robotics and Biomimetics (ROBIO).December 4-9,2023.Koh Samui,Thailand.IEEE,2023:1-6.
[7]
白强.机器视觉中语义分割及目标抓取的深度学习技术研究[D].贵阳:贵州大学,2022.
[8]
BaiQiang.Research on deep learning techniques for semantic segmentation and target grasping in machine vision[D].Guiyang:Guizhou University,2022.
ZangChunhua, ZhouJieqi, LiuGuixiong.Research on recognition method of mechanical parts based on Mask R-CNN instance segmentation[J].Electronic Measurement Technology,2021,44(22):32-36.
ZhaiWeifeng, LuWentao, XueTonglai.Robotic arm gripping position detection method based on instance segmentation[J].Instrument Technique and Sensor,2024(10):89-94,116.
[13]
李乐.面向非结构化任务场景的机器人目标识别及定位抓取研究[D].杭州:浙江大学,2023.
[14]
LiLe.Research on robot object recognition and positioning grasp for unstructured task scenarios[D].Hangzhou:Zhejiang University,2023.
[15]
区树晖.基于合成数据和深度学习的管件无序抓取研究[D].广州:华南理工大学,2024.
[16]
QuShuhui.Research on unordered pipe component grasping based on synthetic data and deep learning[D].Guangzhou:South China University of Technology,2024.
[17]
KhanamR, HussainM.YOLOv11:an overview of the key architectural enhancements[J].arXiv.2024. 10.48550/arXiv.2410.17725
WangA, ChenH, LinZ J,et al.Rep ViT:revisiting mobile CNN from ViT perspective[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).June 16-22,2024,Seattle,WA,USA.IEEE,2024:15909-15920.
[24]
LiuW Z, LuH, FuH T,et al.Learning to upsample by learning to sample[C]//2023 IEEE/CVF International Conference on Computer Vision (ICCV).October 1-6,2023.Paris,France.IEEE,2023:6004-6014.
[25]
WangJ Q, ChenK, XuR,et al.CARAFE:content-aware ReAssembly of features[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV).October 27-November 2,2019.Seoul,Korea.IEEE,2019:3007-3016.
[26]
PaszkeA, GrossS, MassaF,et al.PyTorch:an imperative style,high-performance deep learning library[J].arXiv.2019. 10.48550/arXiv.1912.01703
[27]
HuybrechsD, TrefethenL N.Sigmoid functions and multiscale resolution of singularities[J].arXiv.2023. 10.48550/arXiv.2303.01967
[28]
ZhangH, ZhangS.Shape-IoU:more accurate metric considering bounding box shape and scale[J].arXiv.2023. 10.48550/arXiv.2312.17663
XingYuanxiu, LiuZhuanwei, XingYufeng,et al.BDD-DETR:an efficient algorithm for detecting small surface defects on lithium batteries[J].Energy Storage Science and Technology,2025,14(1):370-379.
ChenPengyu, WangLie, LiangYuman,et al.AFL-YOLO:an improved small object detection algorithm based on YOLOv8[J/OL].Telecommunication Engineering,2025:1-15[2025-01-21].
YangHongdan, FuGui, ShaoHuichao,et al.Small object detection in aerial imagery using multi-scale hiearchical feature fusion based approach[J].Computer Engineering and Applications,2025,61(9):230-241.