Computer vision-based tile counting model for automated progress monitoring

Abstract

Construction industry in Korea has grown based on apartments. Tiles are installed in most of bathrooms of apartment houses so the numbers of tiles are huge to manage. Therefore, this paper proposes a computer vision-based model for automated tile counting. This model uses line detection and rectangle packing algorithm. The results of proposed model appeared that the average of accuracy in line detection and intersection points extraction stage is 98.8% with standard deviation of 1.2%. In calculation stage, the proposed model shows 90.1% of accuracy with standard deviation of 9.4% This implies model can solve the overcounting problem by quickly and accurately calculating the number of tiles with images acquired by the inspector's mobile phone in the field. Furthermore, it can contribute to creating a transparent construction industry.