[1]刘 通,管声启,刘懂懂,等.基于改进 YOLO6D 的工业零件位姿检测算法[J].机械与电子,2023,41(11):22-27.
　LIU Tong,GUAN Shengqi,LIU Dongdong,et al.Industrial Parts Pose Detection Algorithm Based on Improved YOLO6D[J].Machinery & Electronics,2023,41(11):22-27.
点击复制

基于改进 YOLO6D 的工业零件位姿检测算法()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
41
期数:
2023年11期
页码:
22-27
栏目:
自动控制与检测
出版日期:
2023-11-23

文章信息/Info

Title:
Industrial Parts Pose Detection Algorithm Based on Improved YOLO6D
文章编号:
1001-2257 ( 2023 ) 11-0022-06
作者:
刘 通; 管声启; 刘懂懂; 张理博
西安工程大学机电工程学院,陕西 西安 710048
Author(s):
LIU Tong ;  GUAN Shengqi ;  LIU Dongdong ;  ZHANG Libo
( School of Mechanical and Electrical Engineering , Xi ’an Polytechnic University , Xi ’an 710048 , China )
关键词:
工业零件; 改进 YOLO6D ;  6D 位姿检测; 关键点定位
Keywords:
industrial parts ;  improved YOLO6D ;  6D pose detection ;  key points positioning
分类号:
TP391.4 ; TH161.1
文献标志码:
A
摘要:
针对工业现场散乱堆叠、零件位姿难以准确检测的问题,提出一种基于改进 YOLO6D 的工业零件位姿检测算法。首先,对原始YOLO6D 网络结构进行改进。以 Darknet53 作为主干网络,并在其残差块内部引入坐标注意力机制,强化神经网络对坐标信息的表达能力;利用空洞空间金字塔池化捕获多尺度上下文信息,实现底层坐标信息与高层语义信息的特征融合;采用 Mish 函数作为激活函数,增强神经网络的鲁棒性。在此基础上,采用改进 YOLO6D 网络检测工业零件 6D 位姿。以工业零件图像作为改进 YOLO6D 网络的输入,直接回归输出目标零件 3D 边界框的 9 个关键点,利用 2D 3D 空间中的映射关系,采用 PnP 算法计算目标零件的 6D 位姿。最后,进行工业零件位姿检测实验验证。实验结果表明,所提算法具有较高的准确性和鲁棒性,为三维空间内工业零件的位姿检测提供了一种有效的思路。
Abstract:
A pose detection algorithm for industrial parts based on improved YOLO6D is proposed to address the issue of difficulty in accurately detecting the pose of scattered and stacked parts in industrial sites.Firstly , the original YOLO6D network structure is improved.Darknet53 is used as the backbone network and the coordinate attention mechanism is introduced within its residual block to enhance the neural network’s ability to express coordinate information ; the multi-scale context information is captured by adding the atrous spatial pyramid pooling , and the feature fusion of the low-level coordinate information and the high-level semantic information is realized ; the Mish function is used as the activation function to enhance the robustness of the neural network.On this basis , the improved YOLO6D network is used to detect the 6D pose of industrial parts.Using industrial part images as the input of the improved YOLO6D network , the 9 key points of the 3D bounding box of the target part are directly regressed and output.Based on the mapping relationship in 2D-3D space , the PnP algorithm is used to calculate the 6D pose of the target part.Finally , experimental verification is conducted on the pose detection of industrial parts.The experimental results show that the proposed algorithm has high accuracy and robustness , providing an effective approach for pose detection of industrial parts in three-dimensional space.

参考文献/References:

[ 1 ] 杨乐,甘忠,李义龙,等 . 基于改进卷积神经网络的零件识别[ J ] . 仪表技术与传感器, 2022 ( 5 ): 82-87.

[ 2 ] 倪奕棋,管声启,管宇灿,等 . 基于改进的 SSD 深度学习算法的双目视觉纱筒识别定位[ J ] . 纺织高校基础科学学报,2021 , 34 ( 2 ): 59-66.
[ 3 ] 任军胜,晁永生 . 基于改进 YOLOX-S 的机械臂自主识别抓取[ J ] . 组合机床与自动化加工技术, 2022 ( 10 ):59-63.
[ 4 ] 吴庆华,朱思斯,周阳,等 . 基于随机采样一致性的规则棒料位姿识别[ J ] . 传感器与微系统, 2019 , 38 ( 9 ): 137-140.
[ 5 ] 翟敬梅,黄乐 . 堆叠散乱目标的 6D 位姿估计和无序分拣[ J ] . 哈尔滨工业大学学报, 2022 , 54 ( 7 ): 136-142.
[ 6 ] 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.
[ 7 ] 陈海永,李龙腾,陈鹏,等 . 复杂场景点云数据的 6D 位姿估计深度学习网络[ J ] . 电子与信息学报,2022 , 44( 5 ): 1591-1601.
[ 8 ] FAN J M , LI S F , ZHENG P , et al.A high-resolution network-based approach for 6D pose estimation of industrial parts [ C ] ∥2021 IEEE 17th International Conference on Automation Science and Engineering ( CASE ) .New York : IEEE , 2021 : 1452-1457.
[ 9 ] 王神龙,雍宇,吴晨睿 . 基于伪孪生神经网络的低纹理工业零件 6D 位姿估计[ J ] . 电子学报, 2023 , 51 ( 1 ): 192-201.
[ 10 ] CHEN C , JIANG X , MIAO S , et al.Texture less shiny objects grasping in a single RGB image using synthetic training data [ J ] .Applied sciences , 2022 , 12 ( 12 ): 6188-6198.
[ 11 ] 李雨龙,陈松,李鑫,等 . 基于深度学习的管件识别与位姿估计研究 [ J ] . 制造技术与机床,2022 ( 12 ):70-75.
[ 12 ] 杨韧韧,吴炳龙,彭晋民,等 . 面向机器人抓取位姿估计网络研究 [ J / OL ] . 机械科学与技术 ( 2023-03-17 )[ 2023 05 20 ] .https :// doi.org / 10.13433 /j.cnki.1003 8728.20230132.?
[ 13 ] TEKIN B , SINHA S N , FUA P.Real-time seamless single shot 6d object pose prediction [ C ]// 2018 IEEE / CVF Conference on Computer Vision and Pattern Recognition.New York : IEEE , 2018 : 292-301.
[ 14 ] LI S Q , XU C , XIE M.A robust O ( n ) solution to the perspective-n-point problem [ J ] .IEEE Transactions on pattern analysis and machine intelligence , 2012 , 34 ( 7 ): 1444-1450.
[ 15 ] HOU Q B , ZHOU D Q , FENG J S.Coordinate attention for efficient mobile network design [ C ]// 2021 IEEE / CVF Conference on Computer Vision and Pattern Recognition( CVPR ) .New York : IEEE , 2021 :13713-13722.
[ 16 ] CHEN L C , PAPANDREOU G , KOKKINOS I , 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 , 2017 , 40 ( 4 ): 834-84.
[ 17 ] MISRA D.Mish : A Self Regularized Non-Monotonic Activation Function [ C ]// The 31st British Machine Vision Conference , 2020 : 1-13.
[ 18 ] BRACHMANN E , MICHEL F , KRULL A , et al.Uncertainty-driven 6d pose estimation of objects and scenes from a single rgb image [ C ]// 2016 IEEE Conference on Computer Vision and Pattern Recognition( CVPR ) .New York : IEEE , 2016 : 3364-3372.
[ 19 ] HE YS , SUN W , HUANG H B , et al.PVN3D : a deep point-wise 3D keypoints voting network for 6DoF pose estimation [ C ]// 2020 IEEE / CVF Conference on Computer Vision and Pattern Recognition( CVPR ) . New York : IEEE , 2020 : 11632-11641.
[ 20 ] SHOTTON J , GLOCKER B , ZACH C , et al.Scene coordinate regression forests for camera relocalization in RGB-D images [ C ]// 2013 IEEE Conference on Computer Vision and Pattern Recognition.New York : IEEE , 2013 : 2930-2937.

备注/Memo

备注/Memo:
收稿日期: 2023-05-31
基金项目:西安市创新能力强基计划 人工智能技术攻关项目( 21RGZN0021 )
作者简介:刘 通 ( 2000- ),男,河北秦皇岛人,硕士研究生,研究方向为机器人视觉;管声启 ( 1971- ),男,安徽安庆人,博士,教授,研究方向为机械零件质量检测、机器人视觉等,通信作者。
更新日期/Last Update: 2023-12-13