工业流水线生产环境下,多种类小批次中小型航空零件在自动化喷涂作业等过程中,往往面临着零件摆放位姿不固定而难以自动化测量与定位的问题。为实现综合满足喷涂效率与精度的零件定位,应用基于自适应启发改进的粒子群算法,随配准情况自适应地采用不同启发方案与尺度,实现了较强后景影响下快速地点云配准。从全局搜索与局部搜索两方面进行优化,在应用随配准状态自适应的学习系数与惯性系数基础上,引入自适应尺度的滞步系数与类梯度下降运算,解决粒子群算法收敛速度缓慢与易陷入局部最优问题。基于正态分布置信度准则剔除适应度离群值降低后景点云对配准的影响,实现分割后点云与理论数模的智能配准,获得零件摆放的精确位姿。应用二维(2D)视觉减少测量时间、提供点云分割的依据,并多线程并发同步运算提高效率。最终实现了在3 min的节拍内完成三维(3D)点云的自动化扫描,以及在平均2.5 min节拍内实现2 mm精度的多种零件自主配准定位。
Abstract
[Objective] In the realm of industrial assembly line production, the automation of spraying operations for small- and medium-sized parts often encounters significant challenges. The primary issue lies in the inconsistent placement of different parts, complicating the automation of measurement and positioning tasks. [Methods] To address this, a sophisticated approach leveraging 2D images has been developed to assist in the preliminary positioning and classification of parts. This method involves planning the scanning and measurement path of a 3D camera based on the preliminary positioning information provided by 2D image processing. However, the cloud data points captured by the 3D camera include a large amount of irrelevant background information. To isolate accurate point cloud data for the parts, the data is segmented according to the preliminary positioning information of the parts. Point cloud registration, a critical challenge within robotics and computer vision fields, involves estimating a rigid transformation to align one point cloud to another. The final step entails registering the measured point cloud with the 3D digital model of the part to determine the exact position and orientation of the part. The 2D image is processed successively by filtering, gray equalization, binarization, morphological processing, and contour extraction. These steps effectively separate the image foreground from the background and identify the parts within the image. Recently, an innovative adaptation of the classic particle swarm optimization algorithm, enhanced with an adaptive heuristic, has been employed to tailor different schemes and scales based on specific registration conditions. This approach achieves rapid location cloud registration under strong background noise. To overcome issues of slow convergence speed and the tendency to settle into local optima, this improved algorithm incorporates learning and inertia coefficients of adaptive state registration, along with stalling coefficients and gradient descent operations for adaptive scaling. Employing a normal distribution confidence criterion minimizes the effect of fitness outliers on registration, facilitating intelligent alignment between the point cloud and the theoretical numerical model. This helps achieve precise determination of the parts' positions and orientations. [Results] Integrating two-dimensional (2D) vision technology significantly reduces measurement times and improves efficiency through multithreaded concurrent synchronization operations. Finally, the automatic scanning of three-dimensional (3D) point clouds and the autonomous registration positioning of various parts are accomplished within 3 min, achieving an average accuracy of 2 mm. When aligning identical point clouds, the algorithm demonstrates a registration accuracy of 0.002 mm. [Conclusions] Despite its robustness under strong back-point cloud influence, the algorithm's registration accuracy is still greatly affected. In addition, the inherent discrepancy between the sampling consistency of the scanned point cloud and the theoretical numerical model introduces an error of about 0.5 mm, limiting further improvements in registration accuracy. Due to their limited features and small size, some aviation parts are easily misidentified and prone to significant attitude registration errors. It is necessary to further combine the advantages of 2D images and 3D point cloud data to enhance the robustness of the identification and positioning of aviation parts.
关键词
点云配准 /
粒子群算法 /
自适应启发 /
自动化定位 /
2D-3D视觉融合
Key words
point cloud registration /
particle swarm optimization /
adaptive heuristics /
automation positioning /
2D-3D vision fusion
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] WANG R Y, TANG L W, TANG T. Fast sample adaptive offset jointly based on HOG features and depth information for VVC in visual sensor networks[J]. Sensors, 2020, 20(23):6754.
[2] 张慧,王坤峰,王飞跃.深度学习在目标视觉检测中的应用进展与展望[J].自动化学报, 2017, 43(8):1289-1305. ZHANG H, WANG K F, WANG F Y. Advances and perspectives on applications of deep learning in visual object detection[J]. Acta Automatica Sinica, 2017, 43(8):1289-1305.(in Chinese)
[3] 赵继,赵军,张雷,等.焊缝磨抛机器人视觉算法实现及其试验研究[J].机械工程学报, 2013, 49(20):42-48. ZHAO J, ZHAO J, ZHANG L, et al. Vision algorithm and experimental study on the robotic weld-bead grinding and polishing system[J]. Journal of Mechanical Engineering, 2013, 49(20):42-48.(in Chinese)
[4] AOKI Y, GOFORTH H, SRIVATSAN R A, et al. PointNet LK:Robust&efficient point cloud registration using pointnet[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Los Alamitos, USA:IEEE, 2019:7163-7172.
[5] LU W X, WAN G W, ZHOU Y, et al. Deep VCP:An end-to-end deep neural network for point cloud registration[C]//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision. Seoul, Korea (South):IEEE, 2019:12-21.
[6] WANG Y, SOLOMON J. Deep closest point:Learning representations for point cloud registration[C]//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision. Seoul, Korea (South):IEEE, 2019:3523-3532.
[7] CHARLES R Q, SU H, KAICHUN M, et al. PointNet:Deep learning on point sets for 3D classification and segmentation[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, USA:IEEE:2017:77-85.
[8] LI J H, ZHANG C H, XU Z Y, et al. Iterative distance-aware similarity matrix convolution with mutual-supervised point elimination for efficient point cloud registration[C]//Proceedings of the 16th European Conference on Computer Vision. Glasgow, UK:Springer, 2020:378-394.
[9] YEW Z J, LEE G H. RPM-Net:Robust point matching using learned features[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, USA:IEEE, 2020:11824-11833.
[10] BESL P J, MCKAY N D. A method for registration of 3-D shapes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(2):239-256.
[11] YANG J L, LI H D, CAMPBELL D, et al. Go-ICP:A globally optimal solution to 3D ICP point-set registration[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(11):2241-2254.
[12] YANG J L, LI H D, and JIA Y D. Go-ICP:Solving 3D registration efficiently and globally optimally[C]//Proceedings of the 2013 IEEE International Conference on Computer Vision. Los Alamitos:IEEE, 2013:1457-1464.
[13] FROME A, HUBER D, KOLLURI R, et al. Recognizing objects in range data using regional point descriptors[C]//Proceedings of the 8th European Conference on Computer Vision. Prague, Czech Republic:Springer, 2004:224-237.
[14] RUSU R B, BLODOW N, MARTON Z C, et al. Aligning point cloud views using persistent feature histograms[C]//Proceedings of 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems. Nice, France:IEEE, 2008:3384-3391.
[15] RUSU R B, BLODOW N, BEETZ M. Fast point feature histograms (FPFH) for 3D registration[C]//Proceedings of 2009 IEEE International Conference on Robotics and Automation. Kobe, Japan:IEEE, 2009:3212-3217.
[16] YANG Y, Chen W L, Wang M Y, et al. Color point cloud registration based on supervoxel correspondence[J]. IEEE Access, 2020,8:7362-7372
[17] JUNG S, SONG S, CHANG M, et al.. Range image registration based on 2D synthetic images[J]. Computer-Aided Design, 2018. 94:16-27.
[18] ZENG A, SONG S R, NIEBNER M, et al.2017. 3D match:Learning local geometric descriptors from RGB-D reconstructions[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, USA:IEEE.2017:199-208.
[19] E A, SEKIKAWA Y,ISHIKAWA K, et al. CorsNet:3D point cloud registration by deep neural network[J]. IEEE Robotics and Automation Letters, 2020(3):3960-3966.
[20] KENNEDY J, EBERHART R. Particle swarm optimization[C]//Proceedings of the ICNN'95-International Conference on Neural Networks. Perth, Australia:IEEE, 1995:1942-1948.
[21] GE Y Q, WANG B Y, NIE J H, et al. A point cloud registration method combining enhanced particle swarm optimization and iterative closest point method[C]//Proceedings of 2016 Chinese Control and Decision Conference. Yinchuan, China:IEEE, 2016:2810-2815.
[22] 韩贤权,朱庆,丁雨淋,等.散乱点云数据精配准的粒子群优化算法[J].武汉大学学报(信息科学版), 2014, 39(10):1214-1220. HAN X Q, ZHU Q, DING Y L, et al. Precise registration of scattered cloud data based on the particle swarm optimization[J]. Geomatics and Information Science of Wuhan University, 2014, 39(10):1214-1220.(in Chinese)
[23] 潘峰,陈杰,甘明刚,等.粒子群优化算法模型分析[J].自动化学报, 2006, 32(3):368-377. PAN F, CHEN J, GAN M G, et al. Model analysis of particle swarm optimizer[J]. Acta Automatica Sinica, 2006, 32(3):368-377.(in Chinese)
[24] 张顶学,关治洪,刘新芝.一种动态改变惯性权重的自适应粒子群算法[J].控制与决策, 2008, 23(11):1253-1257. ZHANG D X, GUAN Z H, LIU X Z. Adaptive particle swarm optimization algorithm with dynamically changing inertia weight[J]. Control and Decision, 2008, 23(11):1253-1257.(in Chinese)
[25] 王俊伟,汪定伟.一种带有梯度加速的粒子群算法[J].控制与决策, 2004(11):1298-1300, 1304. WANG J W, WANG D W. Particle swarm optimization algorithm with gradient acceleration[J]. Control and Decision, 2004(11):1298-1300, 1304.(in Chinese)
PDF(15875 KB)
