配准是两个点云之间的刚体变换估计问题,近年来有众多深度学习方法被提出以解决该问题。然而,这些方法的训练集往往来自费时费力且数据量不足的真实采集和人工标定。该问题有望通过渲染技术解决。立足于6D位姿估计应用场景,设计了一套基于Point Transformer深度学习点云特征提取网络的点云配准算法流程。应用渲染技术将ShapeNet数据集模型在不同视角生成大量局部投影,构建训练集。使用InfoNCE作为损失函数,以对比学习的方法使Point Transformer网络学习点云的逐点特征描述。开发了Rendering-ICP算法以优化精配准环节,作为最终位姿估计结果。在Linemod数据集上验证本套算法流程,使用VSD回归作为评价指标,与现存方法对比,获得了7%的性能提升。
Abstract
Alignment is the problem of estimating the rigid body transformation between two point clouds, and numerous deep learning methods have been proposed to solve this problem in recent years.However, the training sets of these methods often come from time-consuming and insufficient amount of data for real acquisition and manual calibration.This problem is expected to be solved by rendering techniques.A point cloud alignment algorithm process based on Point Transformer deep learning point cloud feature extraction network is designed for 6D pose estimation application scenario.The rendering technique is applied to generate a large number of local projections of the ShapeNet dataset model in different viewpoints to construct the training set.Using InfoNCE as the loss function, the Point Transformer network is made to learn the point-by-point feature description of point clouds with a contrast learning approach.The Rendering-ICP algorithm is developed to optimize the fine collocation link as the final bit-pose estimation result.The present algorithmic procedure is validated on the Linemod dataset, using VSD regression as an evaluation metric, and a 7% performance improvement is obtained compared to the extant method.
关键词
计算机视觉 /
三维点云 /
渲染 /
深度学习 /
点云配准
{{custom_keyword}} /
Key words
computer vision /
3D point cloud /
rendering /
deep learning /
point cloud registration
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] LIANG Z F, GUO Y L, FENG Y L, et al.Stereo matching using multi-level cost volume and multi-scale feature constancy[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 43(1):300-315.
[2] GUO Y, BENNAMOUN M, SOHEL F, et al.3D object recognition in cluttered scenes with local surface features:a survey[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 36(11):2270-2287.
[3] HUANG X S, MEI G F, ZHANG J, et al.A comprehensive survey on point cloud registration[J].ArXiv Preprint ArXiv, 2021(2103):1-12.
[4] CHEN X Z, MA H M, WAN J, et al.Multi-view 3D object detection network for autonomous driving[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017.
[5] CHENG L, CHEN S, LIU X Q, et al.Registration of laser scanning point clouds:a review[J].Sensors, 2018, 18(5):1641-1650.
[6] BESL P J, MCKAY H D.A method for registration of 3D shapes[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 1992, 14(2):239-256.
[7] YANG C M, MEDIONI G.Object modeling by registration of multiple range images[J].Image and Vision Computing, 1992, 10(3):145-155.
[8] BRENNER C, DOLD C, RIPPERDA N.Coarse orientation of terrestrial laser scans in urban environments[J].Isprs Journal of Photogrammetry & Remote Sensing, 2008, 63(1):4-18.
[9] SEGAL A, HHNEL D, THRUN S.Generalized-ICP[C]//Robotics:Science and Systems V, University of Washington, Seattle, 2009.
[10] MACH C.Random sample consensus:a paradigm for model fitting with application to image analysis and automated cartography[J].ACM, 1981(24):381-395.
[11] YANG H, SHI J, CARLONE L.Teaser:fast and certifiable point cloud registration[J].IEEE Transactions on Robotics, 2020, 37(2):314-333.
[12] RUSU R B, BLODOW N, BEETZ M.Fast point feature histograms (FPFH) for 3D registration[C]//2009 IEEE International Conference on Robotics and Automation.IEEE, 2009.
[13] DROST B, ULRICH M, NAVAB N, et al.Model globally, match locally:efficient and robust 3D object recognition[C]//Computer Vision & Pattern Recognition.IEEE, 2010.
[14] ZENG A, XIAO J R.3D match:learning local geometric descriptors from RGB-D reconstructions[C]//Conference on Computer Vision and Pattern Recognition (CVPR).IEEE, 2017.
[15] DENG H W, BIRDAL T, ILIC S.PPFNet:global context aware local features for robust 3D point matching[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition I, 2018.
[16] CHOY C, PARK J, KOLTUN V.Fully convolutional geometric features[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019.
[17] ZHAO H, JIANG L, JIA J, et al.Point transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021.
[18] CHOY C, DONG W, KOLTUN V.Deep global registration[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020.
[19] YUAN W, ECKART B, KIM K, et al.Deepgmr:learning latent gaussian mixture models for registration[C]//European Conference on Computer Vision.Springer, Cham, 2020.
[20] HUANG X S, MEI G F, ZHANG J.Feature-metric registration:a fast semi-supervised approach for robust point cloud registration without correspondences[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020.
[21] SAITO T, TAKAHASHI T.Comprehensible rendering of 3D shapes[C]//Proceedings of the 17th Annual Conference on Computer Graphics and Interactive Techniques, 1990.
[22] HODAN T, VINEET V, GAL R, et al.Photorealistic image synthesis for object instance detection[C]//International Conference on Image Processing (ICIP).IEEE, 2019.
[23] OORD A, LI Y, VINYALS O.Representation learning with contrastive predictive coding[J].ArXiv Preprint 1807, 2018(1807):1-12.
[24] CHANG A X, FUNKHOUSER T, GUIBAS L, et al.Shapenet:an information-rich 3D model repository[J].ArXiv Preprint, 2015(1512):1-11.
[25] HE K, GKIOXARI G, DOLLAR P, et al.Mask R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017.
[26] HINTERSTOISSER S, LEPETIT V, ILIC S, et al.Model based training, detection and pose estimation of texture-less 3D objects in heavily cluttered scenes[C]//Asian Conference on Computer Vision.Springer, Berlin, Heidelberg, 2012.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(1880 KB)
