结合多尺度体素和高阶条件随机场的点云分类
|
摘要
针对利用经典高阶条件随机场模型进行点云分类时,由于海量节点和无向边导致的点云分类效率低的问题,提出一种结合多尺度体素和高阶条件随机场的点云分类方法.首先以多尺度体素代替海量离散点云作为无向图图模型节点,减少节点和无向边的数量;然后使用超体分割结果作为高阶团,并基于此设计了一种非监督分布性空间上下文作为高阶团特征向量,用于改善分类结果;最后结合构建的图模型和各阶特征向量,采用经典高阶条件随机场模型实现点云数据的自动分类.采用Oakland标准数据集作为实验数据,实验结果表明,该方法在有效地保证分类精度的前提下,高阶条件随机场点云分类模型的分类效率提高了5～10倍.
Aiming at solving the problem of low efficiency of point cloud classification caused by massive nodes and undirected edges when using traditional high-order conditional random field model, a point cloud classification method based on multi-scale voxel and high order random fields is proposed. Firstly, multiscale voxel is utilized as a node of undirected graph to replace the mass of discrete point clouds and reduce the number of nodes and undirected edges. Then, the supervoxel segmentation result is used as a higher-order cluster based on which an unsupervised distributed spatial context is designed as higher-order cluster eigenvector to improve the classification result. Finally, combined with the constructed graph model and each order eigenvector, classical high-order conditional random field model is implemented for automatic point cloud data classification. The Oakland standard dataset is used as the experimental data. Experimental results show that the classification efficiency of the high-order conditional random field point cloud classification model is improved by 5 to 10 times under the premise of ensuring the classification accuracy.
Aiming at solving the problem of low efficiency of point cloud classification caused by massive nodes and undirected edges when using traditional high-order conditional random field model, a point cloud classification method based on multi-scale voxel and high order random fields is proposed. Firstly, multiscale voxel is utilized as a node of undirected graph to replace the mass of discrete point clouds and reduce the number of nodes and undirected edges. Then, the supervoxel segmentation result is used as a higher-order cluster based on which an unsupervised distributed spatial context is designed as higher-order cluster eigenvector to improve the classification result. Finally, combined with the constructed graph model and each order eigenvector, classical high-order conditional random field model is implemented for automatic point cloud data classification. The Oakland standard dataset is used as the experimental data. Experimental results show that the classification efficiency of the high-order conditional random field point cloud classification model is improved by 5 to 10 times under the premise of ensuring the classification accuracy.
引文
[1]LiWenning,ZhangAiwu,WangShumin,etal.Hierarchical classification method for terrestrial laser point clouds[J]. JournalofComputer-AidedDesign&ComputerGraphics,2015,27(8):1555-1561(in Chinese)(李文宁,张爱武,王书民,等.地面激光点云阶层式分类方法[J].计算机辅助设计与图形学学报,2015,27(8):1555-1561)
[2]Tan Ben, Zhong Ruofei, Li Qin. Objects classification with vehicle-borne laser scanning data[J]. Journal of Remote Sensing,2012, 16(1):50-66(in Chinese)(谭贲,钟若飞,李芹.车载激光扫描数据的地物分类方法[J].遥感学报, 2012, 16(1):50-66)
[3]SernaA,MarcoteguiB.Urbanaccessibilitydiagnosisfrom mobilelaserscanningdata[J].ISPRSJournalofPhotogrammetry and Remote Sensing, 2013, 84:23-32
[4]Dong Zhen, Yang Bisheng. Hierarchical extraction of multiple objects from mobile laser scanning data[J]. Acta Geodaetica et Cartographica Sinica, 2015, 44(9):980-987(in Chinese)(董震,杨必胜.车载激光扫描数据中多类目标的层次化提取方法[J].测绘学报, 2015, 44(9):980-987)
[5]Yang B S, Dong Z, Liu Y, et al. Computing multiple aggregation levels and contextual features for road facilitiesrecognitionusingmobilelaserscanningdata[J].ISPRSJournalof Photogrammetry and Remote Sensing, 2017, 126:180-194
[6]Yang B S, Dong Z, Zhao G, et al. Hierarchical extraction of urban objects from mobile laser scanning data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 99:45-57
[7]Wang Z, Zhang L Q, Fang T, et al. A multiscale and hierarchical feature extraction method for terrestrial laser scanning point cloudclassification[J].IEEETransactionsonGeoscienceand Remote Sensing, 2015, 53(5):2409-2425
[8]LiHongjun,LiuXinying,ZhangXiaopeng,etal.A semi-automatic 3D point cloud classification method based on theprobabilitymixtureoflocalshapefeature[J].Journalof Zhejiang University:Science Edition, 2017, 44(1):1-9(in Chinese)(李红军,刘欣莹,张晓鹏,等.局部形状特征概率混合的半自动三维点云分类[J].浙江大学学报:理学版,2017,44(1):1-9)
[9]Guo Bo, Huang Xianfeng, Zhang Fan, et al. Points cloud classification using JointBoost combined with contextual information for feature reduction[J]. Acta Geodaetica et Cartographica Sinica, 2013, 42(5):715-721(in Chinese)(郭波,黄先锋,张帆,等.顾及空间上下文关系的JointBoost点云分类及特征降维[J].测绘学报,2013,42(5):715-721)
[10]Zhang Aiwu, Li Wenning, Duan Yihao, et al. Point cloud classificationbasedonpointfeaturehistogram[J].Journalof Computer-AidedDesign&ComputerGraphics,2016,28(5):795-801(in Chinese)(张爱武,李文宁,段乙好,等.结合点特征直方图的点云分类方法[J].计算机辅助设计与图形学学报,2016,28(5):795-801)
[11]Weinmann M, Jutzi B, Mallet C. Semantic 3D scene interpretation:a framework combining optimal neighborhood size selectionwithrelevantfeatures[J].ISPRSAnnalsofPhotogrammetry, Remote Sensing and Spatial Information Sciences, 2014,II-3:181-188
[12]Weinmann M, Schmidt A, Mallet C, et al. Contextual classificationofpointclouddatabyexploitingindividual3Dneigbourhoods[J].ISPRSAnnalsofPhotogrammetry,Remote SensingandSpatialInformationSciences,2015,II-3(W4):271-278
[13]Niemeyer J, Rottensteiner F, Soergel U. Contextual classification of lidar data and building object detection in urban areas[J].ISPRS Journal of Photogrammetry and Remote Sensing, 2014,87:152-165
[14]Hu H Z, Munoz D, Andrew Bagnell J, et al. Efficient 3-D scene analysis from streaming data[C]//Proceedings of the IEEE InternationalConferenceonRoboticsandAutomation.Los Alamitos:IEEE Computer Society Press, 2013:2297-2304
[15]He Elong, Wang Hongping, Chen Qi, et al. An improved contextual classification method of point cloud[J]. Acta Geodaetica et Cartographica Sinica, 2017, 46(3):362-370(in Chinese)(何鄂龙,王红平,陈奇,等.一种改进的空间上下文点云分类方法[J].测绘学报, 2017, 46(3):362-370)
[16]Triebel R, Schmidt R, MozosóM, et al. Instance-based AMN classification for improved object recognition in 2D and 3D laser range data[C]//Proceedings of the 20th International Joint ConferenceonArtificalIntelligence.SanFrancisco:Morgan Kaufmann Publishers, 2007:2225-2230
[17]Papon J, Abramov A, Schoeler M, et al. Voxel cloud connectivitysegmentation-supervoxelsforpointclouds[C]//Proceedings of the IEEE Computer Vision and Pattern Recognition. Los Alamitos:IEEE Computer Society Press, 2013:2027-2034
[18]Weinmann M, Jutzi B, Mallet C. Feature relevance assessment for the semantic interpretation of 3D point cloud data[J]. ISPRS AnnalsofthePhotogrammetry,RemoteSensingandSpatial Information Sciences, 2013, II-5(W2):313-318
[19]MunozD,AndrewBagnellJ,VandapelN,etal.Contextual classification with functional max-margin Markov networks[C]//Proceedings of the IEEE Computer Vision and Pattern Recognition. Los Alamitos:IEEE Computer Society Press, 2009:975-982
[20]XiongXH,MunozD,AndrewBagnellJ,etal.3-Dscene analysis via sequenced predictions over points and regions[C]//Proceedings of the IEEE International Conference on RoboticsandAutomation.LosAlamitos:IEEEComputerSociety Press, 2011:2609-2616
[21]Munoz D, Vandapel N, Hebert M. Onboard contextual classificationof3-Dpointcloudswithlearnedhigh-orderMarkov random fields[C]//Proceedings of the IEEE International ConferenceonRoboticsandAutomation.LosAlamitos:IEEE Computer Society Press, 2009:2009-2016
[22]Boykov Y, Veksler O, Zabih R. Fast approximate energy minimization via graph cuts[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 23(11):1222-1239
[2]Tan Ben, Zhong Ruofei, Li Qin. Objects classification with vehicle-borne laser scanning data[J]. Journal of Remote Sensing,2012, 16(1):50-66(in Chinese)(谭贲,钟若飞,李芹.车载激光扫描数据的地物分类方法[J].遥感学报, 2012, 16(1):50-66)
[3]SernaA,MarcoteguiB.Urbanaccessibilitydiagnosisfrom mobilelaserscanningdata[J].ISPRSJournalofPhotogrammetry and Remote Sensing, 2013, 84:23-32
[4]Dong Zhen, Yang Bisheng. Hierarchical extraction of multiple objects from mobile laser scanning data[J]. Acta Geodaetica et Cartographica Sinica, 2015, 44(9):980-987(in Chinese)(董震,杨必胜.车载激光扫描数据中多类目标的层次化提取方法[J].测绘学报, 2015, 44(9):980-987)
[5]Yang B S, Dong Z, Liu Y, et al. Computing multiple aggregation levels and contextual features for road facilitiesrecognitionusingmobilelaserscanningdata[J].ISPRSJournalof Photogrammetry and Remote Sensing, 2017, 126:180-194
[6]Yang B S, Dong Z, Zhao G, et al. Hierarchical extraction of urban objects from mobile laser scanning data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 99:45-57
[7]Wang Z, Zhang L Q, Fang T, et al. A multiscale and hierarchical feature extraction method for terrestrial laser scanning point cloudclassification[J].IEEETransactionsonGeoscienceand Remote Sensing, 2015, 53(5):2409-2425
[8]LiHongjun,LiuXinying,ZhangXiaopeng,etal.A semi-automatic 3D point cloud classification method based on theprobabilitymixtureoflocalshapefeature[J].Journalof Zhejiang University:Science Edition, 2017, 44(1):1-9(in Chinese)(李红军,刘欣莹,张晓鹏,等.局部形状特征概率混合的半自动三维点云分类[J].浙江大学学报:理学版,2017,44(1):1-9)
[9]Guo Bo, Huang Xianfeng, Zhang Fan, et al. Points cloud classification using JointBoost combined with contextual information for feature reduction[J]. Acta Geodaetica et Cartographica Sinica, 2013, 42(5):715-721(in Chinese)(郭波,黄先锋,张帆,等.顾及空间上下文关系的JointBoost点云分类及特征降维[J].测绘学报,2013,42(5):715-721)
[10]Zhang Aiwu, Li Wenning, Duan Yihao, et al. Point cloud classificationbasedonpointfeaturehistogram[J].Journalof Computer-AidedDesign&ComputerGraphics,2016,28(5):795-801(in Chinese)(张爱武,李文宁,段乙好,等.结合点特征直方图的点云分类方法[J].计算机辅助设计与图形学学报,2016,28(5):795-801)
[11]Weinmann M, Jutzi B, Mallet C. Semantic 3D scene interpretation:a framework combining optimal neighborhood size selectionwithrelevantfeatures[J].ISPRSAnnalsofPhotogrammetry, Remote Sensing and Spatial Information Sciences, 2014,II-3:181-188
[12]Weinmann M, Schmidt A, Mallet C, et al. Contextual classificationofpointclouddatabyexploitingindividual3Dneigbourhoods[J].ISPRSAnnalsofPhotogrammetry,Remote SensingandSpatialInformationSciences,2015,II-3(W4):271-278
[13]Niemeyer J, Rottensteiner F, Soergel U. Contextual classification of lidar data and building object detection in urban areas[J].ISPRS Journal of Photogrammetry and Remote Sensing, 2014,87:152-165
[14]Hu H Z, Munoz D, Andrew Bagnell J, et al. Efficient 3-D scene analysis from streaming data[C]//Proceedings of the IEEE InternationalConferenceonRoboticsandAutomation.Los Alamitos:IEEE Computer Society Press, 2013:2297-2304
[15]He Elong, Wang Hongping, Chen Qi, et al. An improved contextual classification method of point cloud[J]. Acta Geodaetica et Cartographica Sinica, 2017, 46(3):362-370(in Chinese)(何鄂龙,王红平,陈奇,等.一种改进的空间上下文点云分类方法[J].测绘学报, 2017, 46(3):362-370)
[16]Triebel R, Schmidt R, MozosóM, et al. Instance-based AMN classification for improved object recognition in 2D and 3D laser range data[C]//Proceedings of the 20th International Joint ConferenceonArtificalIntelligence.SanFrancisco:Morgan Kaufmann Publishers, 2007:2225-2230
[17]Papon J, Abramov A, Schoeler M, et al. Voxel cloud connectivitysegmentation-supervoxelsforpointclouds[C]//Proceedings of the IEEE Computer Vision and Pattern Recognition. Los Alamitos:IEEE Computer Society Press, 2013:2027-2034
[18]Weinmann M, Jutzi B, Mallet C. Feature relevance assessment for the semantic interpretation of 3D point cloud data[J]. ISPRS AnnalsofthePhotogrammetry,RemoteSensingandSpatial Information Sciences, 2013, II-5(W2):313-318
[19]MunozD,AndrewBagnellJ,VandapelN,etal.Contextual classification with functional max-margin Markov networks[C]//Proceedings of the IEEE Computer Vision and Pattern Recognition. Los Alamitos:IEEE Computer Society Press, 2009:975-982
[20]XiongXH,MunozD,AndrewBagnellJ,etal.3-Dscene analysis via sequenced predictions over points and regions[C]//Proceedings of the IEEE International Conference on RoboticsandAutomation.LosAlamitos:IEEEComputerSociety Press, 2011:2609-2616
[21]Munoz D, Vandapel N, Hebert M. Onboard contextual classificationof3-Dpointcloudswithlearnedhigh-orderMarkov random fields[C]//Proceedings of the IEEE International ConferenceonRoboticsandAutomation.LosAlamitos:IEEE Computer Society Press, 2009:2009-2016
[22]Boykov Y, Veksler O, Zabih R. Fast approximate energy minimization via graph cuts[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 23(11):1222-1239