一种高压输电走廊机载激光点云分类方法
|
摘要
针对输电线路现有点云分类方法存在的分类效率较低及精度不高等问题,该文从高压输电走廊的地物分布特点出发,提出一种基于JointBoost的高压输电走廊点云分类方法。该方法将三维点云转换为二维影像并基于Hough变换在影像上检测输电走廊候选区域;对候选区域每个点定义并计算多尺度局部特征向量,包括高程特征、连通特征、张量特征和平面特征;根据多尺度局部特征用JointBoost分类器将待分类点云分为地面、植被、电力线和电力塔4类。实验数据表明,该方法能有效地减少高压输电走廊的点云处理数量,提高分类效率,且选取的多尺度特征能有效地表达输电走廊内地物的分布特点,具有较高的分类精度。
To solve problems of low efficiency and low precision of existing methods for high voltage transmission corridor classification from LiDAR data,considering objects' distribution in high voltage transmission corridor,this paper presented a Jointboost-based classification method for high voltage transmission corridor from airborne LiDAR point cloud.Firstly,the original LiDAR point cloud were transformed into image,and then the candidate region of power transmission corridor was detected by Hough transformation;secondly,multiscale local features of point cloud in candidate region are defined and calculated,including the feature of height,connection,tensor and plane;finally,according to their local features,the point cloud were classified into four categories by a JointBoost classifier:ground,vegetation,power line and power pylon.Experimental results showed the proposed method could efficiently reduce the number of processing data,and the selected features could well represent the objects' distribution in high voltage transmission corridor,and the point cloud of high voltage transmission corridor were classified with high precision.
To solve problems of low efficiency and low precision of existing methods for high voltage transmission corridor classification from LiDAR data,considering objects' distribution in high voltage transmission corridor,this paper presented a Jointboost-based classification method for high voltage transmission corridor from airborne LiDAR point cloud.Firstly,the original LiDAR point cloud were transformed into image,and then the candidate region of power transmission corridor was detected by Hough transformation;secondly,multiscale local features of point cloud in candidate region are defined and calculated,including the feature of height,connection,tensor and plane;finally,according to their local features,the point cloud were classified into four categories by a JointBoost classifier:ground,vegetation,power line and power pylon.Experimental results showed the proposed method could efficiently reduce the number of processing data,and the selected features could well represent the objects' distribution in high voltage transmission corridor,and the point cloud of high voltage transmission corridor were classified with high precision.
引文
[1]ZHOU Ruqin,JIANG Wanshou,HUANG Wei,et al.A heuristic method for power pylon reconstructionfrom airborne LiDAR data[J].Remote Sensing,2017,9(11):1172.
[2]魏峰.超高压输电线路巡检系统的研究[D].济南:山东大学,2013.(WEI Feng.Research on EHV trans-mission line inspection system[D].Jinan:ShangdongUniversity,2013.)
[3]GUO Bo,LI Qingquan,HUANG Xianfeng,et al.Animproved method for power-line reconstruction frompoint cloud data[J].Remote Sensing,2016,8(1):36.
[4]邵允临,曹晋恩,尚大伟.直升机巡检华北电网超高压输电线路[J].中国电力,2003,36(7):35-38.(SHAOYunlin,CAO Jinen,SHANG Dawei.Patrol inspectionof EHV electric power transmission line with helicopter inNorth China Power Network[J].Electric Power,2003,36(7):35-38.)
[5]周汝琴,翟瑞聪,江万寿,等.机载激光点云数据中分裂导线自动提取和重建[J].测绘科学,2018,43(6):124-130.(ZHOU Ruqin,ZHAI Ruicong,JIANG Wanshou,et al.Automatic extraction and reconstruction of bundleconductors from airborne LiDAR point clouds[J].Scienceof Surveying and Mapping,2018,43(6):124-130.)
[6]郭波,黄先锋,张帆,等.顾及空间上下文关系的Joint-Boost点云分类及特征降维[J].测绘学报,2013,42(5):715-721.(GUO Bo,HUANG Xianfeng,ZHANGFan,et al.Points cloud classification using jointboostcombined with contextual information for featurereduction[J].Acta Geodaetica et Cartographica Sinica,2013,42(5):715-721.)
[7]徐博,刘正军,王坚.基于激光点云数据电力线的提取及安全检测[J].激光杂志,2017,38(7):48-51.(XUBo,LIU Zhengjun,WANG Jian.Extraction and securitydetection of power line based on laser point cloud data[J].Laser Journal,2017,38(7):48-51.)
[8]王平华,习晓环,王成,等.机载激光雷达数据中电力线的快速提取[J].测绘科学,2017,42(2):154-158,171.(WANG Pinghua,XI Xiaohuan,WANG Cheng,et al.Study on power line fast extraction based on airborneLiDAR data[J].Science of Surveying and Mapping,2017,42(2):154-158,171.)
[9]ZHU Lingli,HYYPPJ.Fully-automated power lineextraction from airborne laser scanning point clouds inforest areas[J].Remote Sensing,2014,6(11):11267-11282.
[10]WANG Yanjun,CHEN Qi,LIU Lin,et al.Supervisedclassification of power lines from airborne LiDAR datain urban Areas[J].Remote Sensing.2017,9(8):771.
[11]GUO Bo,HUANG Xianfeng,ZANG Fan,et al.Classi-fication of airborne laser scanning data using Joint-Boost[J].ISPRS Journal of Photogrammetry &Remote Sensing,2015,100,71-83.
[12]马文,耿贞伟,张小波.基于随机森林方法的输电巡线点云自动分类[J].中国科技纵横,2017(19):147-148,151.(MA Wen,GENG Zhenwei,ZHANG Xiaobo.Automatic classification of transmission line inspectionpoint cloud based on random forest method[J].ChinaScience &Technology Overview,2017(19):147-148,151.)
[13]余洁,穆超,冯延明,等.机载LiDAR点云数据中电力线的提取方法研究[J].武汉大学学报(信息科学版),2011,36(11):1275-1279.(YU Jie,MU Chao,FENGYanming,et al.Powerlines extraction techniques fromairborne LiDAR data[J].Geomatics and InformationScience of Wuhan University,2011,36(11):1275-1279.)
[14]LIU Yuee,LI Zhengrong,HAYWARD R,et al.Classi-fication of airborne LiDAR intensity data using statis-tical analysis and hough transform with application topower line corridors[C]∥Proceedings of the 2009Digital Image Computing:Techniques &Applications.Washington DC,USA:IEEE Computer Society,2009:462-467.
[15]李子龙,刘伟铭.基于JointBoost I2C距离度量的图像分类方法[J].华南理工大学学报(自然科学版),2015,43(5):114-119.(LI Zilong,LIU Weiming.Imageclassification based on JointBoost I2Cdistance metric[J].Journal of South China University of Technology(Natural Science Edition),2015,43(5):114-119.)
[16]GUO Yulan,BENNAMOUN M,SOHEL F,et al.3Dobject recognition in cluttered scenes with local surfacefeatures:a survey[J].IEEE Transactions on PatternAnalysis and Machine Intelligence,2014,36(11):2270-2287.
[17]CHEHATA N,GUO Li,MALLET C.Airborne LiDARfeature selection for urban classification using randomforests[C/OL]∥Laser scanning 2009,IAPRS,2009:207-212[2017-12-23].http:∥www.isprs.org/proceedings/XXXVIII/3-W8/papers/207_laserscanning09.pdf.
[18]TORRALBA A,MURPHY K P,FREEMAN W T.Sharing visual features for multiclass and multi-viewobjects detection[J].IEEE Transactions on PatternAnalysis and Machine Intelligence,2007,29(5):854-869.
[19]RATSCH G,ONODA T,MULLER K R.Soft marginsfor AdaBoost[J].Machine Learning,2001,42(3):287-320.
[20]RTSCH G,SCHLKOPF B,MIKA S,et al.SVMand Boosting:one class[EB/OL].[2017-12-23].http:∥pdfs.semanticscholar.org/f4fe/3c22e7f6cc86ee2fd44314d81c4be90efbee.pdf.
[22]于玲,吴铁军.集成学习:BOOSTING算法综述[J].模式识别与人工智能,2004,17(1):52-59.(YU Ling,WU Tiejun.Assemble learning:a survey of Boostingalgorithms[J].PR &AI,2004,17(1):52-59.)
[23]董乐红,耿国华,高原.Boosting算法综述[J].计算机应用于软件,2006,23(8):27-29.(DONG Lehong,GENG Guohua,GAO Yuan.Survey of Boosting[J].Computer Application and Software,2006,23(8):27-29.)
[24]周志华.机器学习[M].北京:清华大学出版社,2016.(ZHOU Zhihua.Machine learning[M].Beijing:Tsinghua University Press,2016.
[2]魏峰.超高压输电线路巡检系统的研究[D].济南:山东大学,2013.(WEI Feng.Research on EHV trans-mission line inspection system[D].Jinan:ShangdongUniversity,2013.)
[3]GUO Bo,LI Qingquan,HUANG Xianfeng,et al.Animproved method for power-line reconstruction frompoint cloud data[J].Remote Sensing,2016,8(1):36.
[4]邵允临,曹晋恩,尚大伟.直升机巡检华北电网超高压输电线路[J].中国电力,2003,36(7):35-38.(SHAOYunlin,CAO Jinen,SHANG Dawei.Patrol inspectionof EHV electric power transmission line with helicopter inNorth China Power Network[J].Electric Power,2003,36(7):35-38.)
[5]周汝琴,翟瑞聪,江万寿,等.机载激光点云数据中分裂导线自动提取和重建[J].测绘科学,2018,43(6):124-130.(ZHOU Ruqin,ZHAI Ruicong,JIANG Wanshou,et al.Automatic extraction and reconstruction of bundleconductors from airborne LiDAR point clouds[J].Scienceof Surveying and Mapping,2018,43(6):124-130.)
[6]郭波,黄先锋,张帆,等.顾及空间上下文关系的Joint-Boost点云分类及特征降维[J].测绘学报,2013,42(5):715-721.(GUO Bo,HUANG Xianfeng,ZHANGFan,et al.Points cloud classification using jointboostcombined with contextual information for featurereduction[J].Acta Geodaetica et Cartographica Sinica,2013,42(5):715-721.)
[7]徐博,刘正军,王坚.基于激光点云数据电力线的提取及安全检测[J].激光杂志,2017,38(7):48-51.(XUBo,LIU Zhengjun,WANG Jian.Extraction and securitydetection of power line based on laser point cloud data[J].Laser Journal,2017,38(7):48-51.)
[8]王平华,习晓环,王成,等.机载激光雷达数据中电力线的快速提取[J].测绘科学,2017,42(2):154-158,171.(WANG Pinghua,XI Xiaohuan,WANG Cheng,et al.Study on power line fast extraction based on airborneLiDAR data[J].Science of Surveying and Mapping,2017,42(2):154-158,171.)
[9]ZHU Lingli,HYYPPJ.Fully-automated power lineextraction from airborne laser scanning point clouds inforest areas[J].Remote Sensing,2014,6(11):11267-11282.
[10]WANG Yanjun,CHEN Qi,LIU Lin,et al.Supervisedclassification of power lines from airborne LiDAR datain urban Areas[J].Remote Sensing.2017,9(8):771.
[11]GUO Bo,HUANG Xianfeng,ZANG Fan,et al.Classi-fication of airborne laser scanning data using Joint-Boost[J].ISPRS Journal of Photogrammetry &Remote Sensing,2015,100,71-83.
[12]马文,耿贞伟,张小波.基于随机森林方法的输电巡线点云自动分类[J].中国科技纵横,2017(19):147-148,151.(MA Wen,GENG Zhenwei,ZHANG Xiaobo.Automatic classification of transmission line inspectionpoint cloud based on random forest method[J].ChinaScience &Technology Overview,2017(19):147-148,151.)
[13]余洁,穆超,冯延明,等.机载LiDAR点云数据中电力线的提取方法研究[J].武汉大学学报(信息科学版),2011,36(11):1275-1279.(YU Jie,MU Chao,FENGYanming,et al.Powerlines extraction techniques fromairborne LiDAR data[J].Geomatics and InformationScience of Wuhan University,2011,36(11):1275-1279.)
[14]LIU Yuee,LI Zhengrong,HAYWARD R,et al.Classi-fication of airborne LiDAR intensity data using statis-tical analysis and hough transform with application topower line corridors[C]∥Proceedings of the 2009Digital Image Computing:Techniques &Applications.Washington DC,USA:IEEE Computer Society,2009:462-467.
[15]李子龙,刘伟铭.基于JointBoost I2C距离度量的图像分类方法[J].华南理工大学学报(自然科学版),2015,43(5):114-119.(LI Zilong,LIU Weiming.Imageclassification based on JointBoost I2Cdistance metric[J].Journal of South China University of Technology(Natural Science Edition),2015,43(5):114-119.)
[16]GUO Yulan,BENNAMOUN M,SOHEL F,et al.3Dobject recognition in cluttered scenes with local surfacefeatures:a survey[J].IEEE Transactions on PatternAnalysis and Machine Intelligence,2014,36(11):2270-2287.
[17]CHEHATA N,GUO Li,MALLET C.Airborne LiDARfeature selection for urban classification using randomforests[C/OL]∥Laser scanning 2009,IAPRS,2009:207-212[2017-12-23].http:∥www.isprs.org/proceedings/XXXVIII/3-W8/papers/207_laserscanning09.pdf.
[18]TORRALBA A,MURPHY K P,FREEMAN W T.Sharing visual features for multiclass and multi-viewobjects detection[J].IEEE Transactions on PatternAnalysis and Machine Intelligence,2007,29(5):854-869.
[19]RATSCH G,ONODA T,MULLER K R.Soft marginsfor AdaBoost[J].Machine Learning,2001,42(3):287-320.
[20]RTSCH G,SCHLKOPF B,MIKA S,et al.SVMand Boosting:one class[EB/OL].[2017-12-23].http:∥pdfs.semanticscholar.org/f4fe/3c22e7f6cc86ee2fd44314d81c4be90efbee.pdf.
[22]于玲,吴铁军.集成学习:BOOSTING算法综述[J].模式识别与人工智能,2004,17(1):52-59.(YU Ling,WU Tiejun.Assemble learning:a survey of Boostingalgorithms[J].PR &AI,2004,17(1):52-59.)
[23]董乐红,耿国华,高原.Boosting算法综述[J].计算机应用于软件,2006,23(8):27-29.(DONG Lehong,GENG Guohua,GAO Yuan.Survey of Boosting[J].Computer Application and Software,2006,23(8):27-29.)
[24]周志华.机器学习[M].北京:清华大学出版社,2016.(ZHOU Zhihua.Machine learning[M].Beijing:Tsinghua University Press,2016.