Determination of On-site Construction Labor Productivity Using Artificial Neural Networks

To enhance the capability of highway bridge construction, an automated on-site construction labor productivity measurement system was developed. Utilizing the advanced technologies of computer vision and the artificial neural network, the developed system first wirelessly acquired a sequence of images of construction labor activities. Then, the human pose analyzing algorithm processed these images in real-time to generate human poses associated with the construction workers at the project site. Next, a portion of the human poses were manually classified into three categories as effective work, ineffective work, and contributory work and were used to train a built-in artificial neural network. Finally, the trained neural network was employed to decide the ongoing worker’s working status by comparing the in-coming images to the developed human poses. As a result, the construction labor productivity was determined from these comparison statistics. The developed system was tested for accuracy on a bridge construction project. The results of the test indicated that the productivity measurements by the neural network were reasonable accurate when compared to the measurements produced by the manual method. This research project made two major contributions to the advancement of construction industry. First, it applied advanced technologies such as computer vision and artificial neural network for analyzing construction operations. Second, the results of this research project made it possible to automatically determine the on-site construction labor productivity in real-time. Thus, engineers and project managers were able to quickly identify on-site labor productivity problems and to take actions immediately to address these problems. Therefore, the success of this research project enhanced the contractors’ capability of managing bridge construction projects.

Language

  • English

Media Info

  • Media Type: DVD
  • Features: Figures; Photos; References; Tables;
  • Pagination: 14p
  • Monograph Title: TRB 89th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

Filing Info

  • Accession Number: 01151222
  • Record Type: Publication
  • Report/Paper Numbers: 10-2481
  • Files: TRIS, TRB
  • Created Date: Jan 25 2010 11:10AM