https://doi.org/10.24928/2023/0108

Inventory and Piling Waste: A Computer Vision Approach

Inshu Chauhan¹, Eelon Lappalainen², Ana Reinbold³, Ilari Palsola⁴ & Olli Seppänen⁵

¹Doctoral Researcher, Department of Civil Engineering, Aalto University, Finland, [email protected], orcid.org/ 0000-0002-3257-2535
²Doctoral Researcher, Department of Civil Engineering, Aalto University, Finland, [email protected], orcid.org/0000-0002-7573-344X
³Doctoral Researcher, Department of Civil Engineering, Aalto University, Finland, [email protected], orcid.org/0002-7774-7984
⁴Project Engineer, Lean and Last Planner in construction, Fimpec PMO Ltd., Finland, [email protected], orcid.org/0000-0001-9779-2144
⁵Associate Professor, Department of Civil Engineering, Aalto University, Finland, [email protected], orcid.org/0000-0002-2008-5924

Abstract

Construction sites contain a lot of waste, and eliminating it enables productivity gains and health and safety improvements. Computer vision is a promising technology that is being used in various construction applications. Construction sites with limited human resources could benefit from automated computer vision-based waste analysis. This paper presents preliminary findings related to the algorithm-based waste detection of piling works and explores potential applications from a visual management perspective. An experimental approach was used in the study, and images from a construction site in Finland were used to train the algorithm. The main findings revealed that the amount of waste shown by the images was substantial and that ground-level and drone images could be combined to create a comprehensive view of pile waste inventories. This paper also presents potential applications of image-based pattern recognition for infrastructure sites where the use of drone and ground-level images is standard practice. Several problems emerged when using transfer learning to train the algorithm, the most significant of which were variations in the scenery of images used for training and the limited number of images. The solutions to these problems lie in collecting more data and experimenting with other deep learning-based methods which will be explored in future.

Keywords

Lean construction, waste, visual management, computer vision, piling.

Files

Reference

Chauhan, I. , Lappalainen, E. , Reinbold, A. , Palsola, I. & Seppänen, O. 2023. Inventory and Piling Waste: A Computer Vision Approach, Proceedings of the 31st Annual Conference of the International Group for Lean Construction (IGLC31) , 435-444. doi.org/10.24928/2023/0108

Download: BibTeX | RIS Format