POINT CLOUD SEGMENTATION USING IMAGE PROCESSING TECHNIQUES FOR STRUCTURAL ANALYSIS

Abstract. Modern surveying techniques, with the combined use of Unmanned Aerial Vehicles (UAV) with low-cost photographic sensors, and photogrammetric techniques, allows obtaining a precise virtual reconstruction of environment and object with centimetre accuracy. Recently, the diffusion of UAV allows the survey of extensive areas significantly reducing survey time and costs. The raw output obtainable from such survey operations consists of a three-dimensional point cloud. Numerous applications in architecture, monitoring and surveying and structural analysis require objects identification in the 3d scene to classify different element in the acquired scene and extract relevant information. Point cloud analysis, and in particular segmentation and classification techniques, are actually used to identify objects within the scenes, assign to a specific class and use them for subsequent studies. These techniques represent an open research theme and the key to add value to the entire process. Actual methodologies are based on 3d spatial analysis on the point cloud. In this paper, starting from photogrammetric reconstruction, a methodology for segmentation and classification of point cloud based on image analysis is presented. The object identification on the image’s dataset is performed using a Neural Network and subsequently the identified object on dataset are transfer into the 3d environment. This classification is performed to segment structural parts of bridges and viaduct, acquire geometric information, and perform a structural analysis to preserve relevant and ancient structure. A case study for the segmentation of the point cloud acquired with an aerial survey of a Viaduct is presented. The performed segmentation allows obtaining structural elements of different type of viaduct and bridges, is propaedeutic to verify the health of the structure and schedule maintenance intervention. The methodology can be applied to different type of bridges, from reinforced concrete to ancient masonry to preserve the state of conservation.

[1]  Fabio Remondino,et al.  SEGMENTATION OF 3D PHOTOGRAMMETRIC POINT CLOUD FOR 3D BUILDING MODELING , 2018, The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences.

[2]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  Anh Nguyen,et al.  3D point cloud segmentation: A survey , 2013, 2013 6th IEEE Conference on Robotics, Automation and Mechatronics (RAM).

[4]  T. Rabbani,et al.  SEGMENTATION OF POINT CLOUDS USING SMOOTHNESS CONSTRAINT , 2006 .

[5]  I. Colomina,et al.  Unmanned aerial systems for photogrammetry and remote sensing: A review , 2014 .

[6]  M. Brook.  Structure from motion in the geosciences , 2017 .

[7]  Fabio Remondino,et al.  UAV PHOTOGRAMMETRY FOR MAPPING AND 3D MODELING - CURRENT STATUS AND FUTURE PERSPECTIVES - , 2012 .

[8]  José García Rodríguez,et al.  A Review on Deep Learning Techniques Applied to Semantic Segmentation , 2017, ArXiv.

[9]  Trevor Darrell,et al.  Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[10]  David G. Lowe,et al.  Object recognition from local scale-invariant features , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[11]  M Lokanath,et al.  Accurate object classification and detection by faster-RCNN , 2017 .

[12]  Alistair A. Young,et al.  Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) , 2017, MICCAI 2017.

[13]  Richard Szeliski,et al.  Modeling the World from Internet Photo Collections , 2008, International Journal of Computer Vision.

[14]  Rob Fergus,et al.  Visualizing and Understanding Convolutional Networks , 2013, ECCV.

[15]  Leonidas J. Guibas,et al.  PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).