|本期目录/Table of Contents|

[1]蒋珺阳,吴晶华*,赵娜娜.基于大核注意力改进的工业管件位姿估计[J].武汉工程大学学报,2024,46(03):304-309.[doi:10.19843/j.cnki.CN42-1779/TQ.202310018]
 JIANG Junyang,WU Jinghua*,ZHAO Nana.Pose estimation of industrial pipe fittings based onlarge kernel attention improvement[J].Journal of Wuhan Institute of Technology,2024,46(03):304-309.[doi:10.19843/j.cnki.CN42-1779/TQ.202310018]
点击复制

基于大核注意力改进的工业管件位姿估计(/HTML)
分享到:

《武汉工程大学学报》[ISSN:1674-2869/CN:42-1779/TQ]

卷:
46
期数:
2024年03期
页码:
304-309
栏目:
机电与信息工程
出版日期:
2024-06-30

文章信息/Info

Title:
Pose estimation of industrial pipe fittings based on
large kernel attention improvement
文章编号:
1674 - 2869(2024)03 - 0304 - 06
作者:
蒋珺阳12吴晶华*2赵娜娜2
1. 安徽建筑大学机械与电气工程学院,安徽 合肥 230009;
2. 中国科学院合肥物质科学研究院智能机械研究所,安徽 合肥 230031
Author(s):
JIANG Junyang1WU Jinghua*2ZHAO Nana2
1. School of Mechanical and Electrical Engineering,Anhui Jianzhu University, Hefei 230009, China;
2. Institute of Intelligent Machines,Hefei Institute of Physical Sciences,Chinese Academy of Sciences,Hefei 230031,China

关键词:
位姿估计密集对应深度学习大核注意力
Keywords:
pose estimation dense correspondence deep learning large kernel attention
分类号:
TP391
DOI:
10.19843/j.cnki.CN42-1779/TQ.202310018
文献标志码:
A
摘要:
在物体六自由度位姿估计任务中,现有算法在真实场景下对具有弱纹理性且摆放存在遮挡的工件难以实现准确的识别。为提高工件识别精度,提出一种基于深度学习改进的位姿估计算法。该算法采用编码器-解码器架构,引入大核注意力组成视觉注意力网络,聚焦不确定性关键点,增强特征提取能力。根据关键点对应构建密集点对关系,求解出候选位姿。实验结果表明,该算法在公共数据集和自建工业管件数据集上识别准确率分别达到了57.4%和62.1%。与高密度表面编码(Surfemb)算法相比准确率分别提升了5.5%和1.9%。这验证了该算法在遮挡场景下有更高的精准度和鲁棒性。

Abstract:
In the task of six-degree-of-freedom pose estimation, the existing algorithms cannot accurately recognize the workpieces with weak texture and occluded placement in real settings. To improve the accuracy of workpiece recognition, an improved pose estimation algorithm based on depth learning was proposed. A large kernel attention was first added to the encoder-decoder architecture to construct a visual attention network so that the network could focus on uncertain key points and enhance feature extraction capability. Then, the candidate pose was obtained according to the key points corresponding to building a dense point-to-point relationship. The experimental results show that the recognition accuracy of the algorithm is 57.4% on the public dataset and 62.1% on the self-built industrial pipe fittings dataset, respectively. Compared with the surface embeddings (Surfemb) algorithm, the accuracy is improved by 5.5% and 1.9%, respectively. This proves that the proposed algorithm has a higher accuracy and robustness in occluded scenes.

参考文献/References:

[1] 郭楠,李婧源,任曦.基于深度学习的刚体位姿估计方法综述[J].计算机科学,2023,50(2):178-189.

[2] 陈立挺,聂晓根.基于双目视觉的机械手识别、定位、抓取系统研究[J].机电工程,2019,36(8):862-866,872.
[3] 张苏沛,刘军,肖澳文,等.基于卷积神经网络的验证码识别[J].武汉工程大学学报,2019,41(1):89-92.
[4] 陈希彤,卢涛.基于全局深度分离卷积残差网络的高效人脸识别算法[J].武汉工程大学学报,2019,41(3):276-282.
[5] HU Y L,HUGONOT J,FUA P,et al. Segmentation-driven 6D object pose estimation[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE,2019:3385-3394.
[6] LI Y,WANG G,JI X Y,et al. DeepIM:deep iterative matching for 6D pose estimation[J].International Journal of Computer Vision,2020,128(7):657-678.
[7] PENG S, LIU Y, HUANG Q X,et al. PVNet:pixel-wise voting network for 6DoF pose estimation[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE,2019:4561-4570.
[8] LEPETIT V,MORENO-NOGUER F,FUA P. EPnP:an accurate o(n) solution to the PnP problem[J]. International Journal of Computer Vision,2009,81(2):155-166.
[9] PARK K, PATTEN T, VINCZE M. Pix2Pose:pixel-wise coordinate regression of objects for 6D pose estimation[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway:IEEE,2019:7668-7677.
[10] 李少飞,史泽林,庄春刚.基于深度学习的物体点云六维位姿估计方法[J].计算机工程,2021,47(8):216-223.
[11] FAN R Z,XU T B,WEI?Z Z.Estimating 6D aircraft pose from keypoints and structures[J].Remote Sensing,2021,13(4):663.
[12] YANG X L,JIA X H,LIANG Y,et al. 6D object pose estimation in cluttered scenes from RGB images[J].Journal of Computer Science and Technology,2022,37(3):719-730.
[13] 王太勇,孙浩文.基于关键点特征融合的六自由度位姿估计方法[J].天津大学学报(自然科学与工程技术版),2022,55(5):543-551.
[14] HAUGAARD R L,BUCH A G. Surfemb:dense and continuous correspondence distributions for object pose estimation with learnt surface embeddings[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE,2022:6749-6758.
[15] LIU S Q,WEI J S,LIU G,et al.?Image classification model based on large kernel attention mechanism and relative position self-attention mechanism[J].?PeerJ Computer Science,2023,9:e1344.
[16] RONNEBERGER O,FISCHER P,BROX T. U-Net:convolutional networks for biomedical image segmentation[C]//Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015. Switzerland:Springer,2015:234-241.

相似文献/References:

备注/Memo

备注/Memo:
收稿日期:2023-10-24
基金项目:安徽省重点实验室基金(IRKL2022KF04);江苏省重点研发计划项目(BE2017001-1)
作者简介:蒋珺阳,硕士研究生。Email:[email protected]
*通信作者:吴晶华,博士,副研究员。Email:[email protected]
引文格式:蒋珺阳,吴晶华,赵娜娜. 基于大核注意力改进的工业管件位姿估计[J]. 武汉工程大学学报,2024,46(3):304-309,316 .

更新日期/Last Update: 2024-07-02