Hierarchical classification of pole‐like objects in mobile laser scanning point clouds. (10th January 2020)
- Record Type:
- Journal Article
- Title:
- Hierarchical classification of pole‐like objects in mobile laser scanning point clouds. (10th January 2020)
- Main Title:
- Hierarchical classification of pole‐like objects in mobile laser scanning point clouds
- Authors:
- Liu, Rufei
Wang, Peng
Yan, Zhaojin
Lu, Xiushan
Wang, Minye
Yu, Jiayong
Tian, Maoyi
Ma, Xinjiang - Abstract:
- Abstract: For the classification of pole‐like objects (trees, lamp posts, traffic lights and traffic signs) in mobile laser scanning (MLS) point clouds, a hierarchical classification method is proposed. The method consists of three major steps. (1) The objects' cylindrical column sections are detected based on the characteristics of arc‐like points using RANSAC after denoising. (2) These detected objects are roughly classified into trees and man‐made poles based on the azimuthal coverage of point clouds above the cylindrical column. (3) Eigenvalue analysis and the principal direction of the upper pole projections are used to differentiate lamp posts, traffic lights and traffic signs. Experimental analysis shows that the method can effectively identify different types of pole‐like objects. Abstract : For the classification of pole‐like objects (trees, lamp posts, traffic lights and traffic signs) in mobile laser scanning (MLS) point clouds, a hierarchical classification method is proposed using three steps. (1) The objects' cylindrical column sections are detected based on their arc‐like characteristics using RANSAC after denoising. (2) These objects are classified into trees and man‐made poles based on the azimuthal coverage of their upper point clouds. (3) Eigenvalue analysis and the principal direction of the upper projections are used to differentiate lamp posts, traffic lights and traffic signs. Résumé: Pour la classification d'objets axiaux (arbres, lampadaires, feux deAbstract: For the classification of pole‐like objects (trees, lamp posts, traffic lights and traffic signs) in mobile laser scanning (MLS) point clouds, a hierarchical classification method is proposed. The method consists of three major steps. (1) The objects' cylindrical column sections are detected based on the characteristics of arc‐like points using RANSAC after denoising. (2) These detected objects are roughly classified into trees and man‐made poles based on the azimuthal coverage of point clouds above the cylindrical column. (3) Eigenvalue analysis and the principal direction of the upper pole projections are used to differentiate lamp posts, traffic lights and traffic signs. Experimental analysis shows that the method can effectively identify different types of pole‐like objects. Abstract : For the classification of pole‐like objects (trees, lamp posts, traffic lights and traffic signs) in mobile laser scanning (MLS) point clouds, a hierarchical classification method is proposed using three steps. (1) The objects' cylindrical column sections are detected based on their arc‐like characteristics using RANSAC after denoising. (2) These objects are classified into trees and man‐made poles based on the azimuthal coverage of their upper point clouds. (3) Eigenvalue analysis and the principal direction of the upper projections are used to differentiate lamp posts, traffic lights and traffic signs. Résumé: Pour la classification d'objets axiaux (arbres, lampadaires, feux de signalisation, panneaux) dans des nuages de points issus de balayage laser mobile (MLS), une méthode de classification hiérarchique est proposée. Cette méthode se divise en trois grandes étapes. (1) Après filtrage du bruit, les sections de colonnes cylindriques sont détectées sur la base des caractéristiques de cibles en forme d'arc grâce à l'algorithme RANSAC. (2) Les objets ainsi détectés sont sommairement classés entre les arbres et les objets artificiels à partir de la couverture azimutale des nuages de points au‐dessus de la colonne cylindrique. (3) L'analyse des valeurs propres et du nombre de directions dominantes dans la partie supérieure du nuage de points, permet de différentier les lampadaires, les feux de signalisation et les panneaux. Une étude expérimentale montre que cette méthode peut effectivement identifier les différents types d'objets axiaux. Zusammenfassung: Zur Klassifzierung von pfahlartigen Objekten wie Bäumen, Lampenmasten, Verkehrsampeln und Verkehrschildern in Punktwolken aus mobilem Laserscanning (MLS) wird eine hierarchische Klassifizierungsmethode vorgeschlagen. Die Methode besteht aus drei Hauptschritten. (1) Die zylindrischen, säulenartigen Abschnitte des Objekts werden, nach einer Rauschminderung, mit dem RANSAC Verfahren auf der Basis der Eigenschaften von Bogenpunkten erkannt. (2) Diese erkannten Objekte werden grob in Bäume und künstliche Pfähle mit Hilfe der horizontalen Abdeckung der Punktwolken über den zylindrischen Säulen klassifiziert. (3) Eine Analyse der Eigenwerte und der Hauptrichtung der oberen Pfahlprojektonen wird genutzt, um Straßenlampen, Verkehrsampeln und Verkehrszeichen zu unterscheiden. Die durchgeführten Experimente zeigen, dass die Methode sehr effektiv die verschiedenen Arten der pfahlartigen Objekte unterscheiden kann. Resumen: Se propone un método jerárquico para la clasificación de objetos tipo poste (árboles, postes de luz, semáforos y señales de tráfico) en nubes de puntos de escaneo láser móvil (MLS). El método consta de tres pasos principales. (1) Los objetos de base cilíndrica se detectan en función de las características de los puntos formando arco utilizando RANSAC después de la eliminación de ruido. (2) Estos objetos detectados se clasifican inicialmente en árboles y postes artificiales según la cobertura azimutal de las nubes de puntos sobre la columna cilíndrica. (3) El análisis del valor propio y la dirección principal de las proyecciones del polo superior se utilizan para diferenciar postes de luz, semáforos y señales de tráfico. El análisis experimental muestra que el método puede identificar efectivamente diferentes tipos de objetos tipo poste. 摘要: 针对车载激光点云中杆状地物(树木,路灯,交通信号灯和交通标志牌)的分类问题,本文提出了一种渐进分类方法。该方法由三个主要部分组成:(1)进行去噪后,根据圆弧状点集的三维特征,使用RANSAC从点云中提取得到杆的柱状部分。(2)基于杆柱状部分上部点云的方位覆盖度,将提取得到的杆状地物粗分类为树木和人造杆状地物。(3)利用特征值分析和杆上部点云在主方向的投影将人造杆状地物分类为路灯、交通信号灯和交通标志牌。实验分析表明,该方法能够有效分类得到不同种类的杆状地物。 … (more)
- Is Part Of:
- Photogrammetric record. Volume 35:Number 169(2020)
- Journal:
- Photogrammetric record
- Issue:
- Volume 35:Number 169(2020)
- Issue Display:
- Volume 35, Issue 169 (2020)
- Year:
- 2020
- Volume:
- 35
- Issue:
- 169
- Issue Sort Value:
- 2020-0035-0169-0000
- Page Start:
- 81
- Page End:
- 107
- Publication Date:
- 2020-01-10
- Subjects:
- eigenvalue analysis -- hierarchical classification -- mobile laser scanning (MLS) -- pole‐like objects -- 3D characteristic statistics -- trees
Photogrammetry -- Periodicals
526.982 - Journal URLs:
- http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1477-9730 ↗
http://onlinelibrary.wiley.com/ ↗ - DOI:
- 10.1111/phor.12307 ↗
- Languages:
- English
- ISSNs:
- 0031-868X
- Deposit Type:
- Legaldeposit
- View Content:
- Available online (eLD content is only available in our Reading Rooms) ↗
- Physical Locations:
- British Library DSC - 6468.100000
British Library DSC - BLDSS-3PM
British Library HMNTS - ELD Digital store - Ingest File:
- 13295.xml