TY - GEN
T1 - Enhanced Appearance-Based Material Classification for the Monitoring of Operation-Level Construction Progress through the Removal of Occlusions
AU - Han, Kevin K.
AU - Muthukumar, Banu
AU - Golparvar-Fard, Mani
N1 - Publisher Copyright:
© ASCE.
PY - 2016
Y1 - 2016
N2 - Tracking and visualizing the status of work-in-progress on construction sites can minimize the gap between short-term and long-term planning and lower coordination costs. Over the past few years, there have been studies on leveraging appearance information in point clouds and 4D building information models (BIM) to automatically detect progress deviations. These methods rely on libraries of construction material images to train the underlying material recognition models. Areas of the images corresponding to BIM elements are extracted and tested with the trained material recognition model. While successful results have been achieved, these extracted images contain occlusions that cause some BIM elements to be misclassified. To improve the accuracy, this paper presents a method that removes occlusions prior to performing material recognition. By creating 3D depth maps and simple linear iterative clustering (SLIC) superpixels, occlusions presented in the images are removed. The presented method is validated with four case studies. The experimental result shows the improved accuracies compared against that of the previous method without occlusion removal.
AB - Tracking and visualizing the status of work-in-progress on construction sites can minimize the gap between short-term and long-term planning and lower coordination costs. Over the past few years, there have been studies on leveraging appearance information in point clouds and 4D building information models (BIM) to automatically detect progress deviations. These methods rely on libraries of construction material images to train the underlying material recognition models. Areas of the images corresponding to BIM elements are extracted and tested with the trained material recognition model. While successful results have been achieved, these extracted images contain occlusions that cause some BIM elements to be misclassified. To improve the accuracy, this paper presents a method that removes occlusions prior to performing material recognition. By creating 3D depth maps and simple linear iterative clustering (SLIC) superpixels, occlusions presented in the images are removed. The presented method is validated with four case studies. The experimental result shows the improved accuracies compared against that of the previous method without occlusion removal.
UR - http://www.scopus.com/inward/record.url?scp=84976415552&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84976415552&partnerID=8YFLogxK
U2 - 10.1061/9780784479827.089
DO - 10.1061/9780784479827.089
M3 - Conference contribution
AN - SCOPUS:84976415552
T3 - Construction Research Congress 2016: Old and New Construction Technologies Converge in Historic San Juan - Proceedings of the 2016 Construction Research Congress, CRC 2016
SP - 879
EP - 889
BT - Construction Research Congress 2016
A2 - Perdomo-Rivera, Jose L.
A2 - Lopez del Puerto, Carla
A2 - Gonzalez-Quevedo, Antonio
A2 - Maldonado-Fortunet, Francisco
A2 - Molina-Bas, Omar I.
PB - American Society of Civil Engineers
T2 - Construction Research Congress 2016: Old and New Construction Technologies Converge in Historic San Juan, CRC 2016
Y2 - 31 May 2016 through 2 June 2016
ER -