Vision-based real-time progress tracking and productivity analysis of the concrete pouring process

Abstract

Regular progress monitoring is essential to keep the project on track consistently, but tracking progress on a daily basis fails to reveal operational issues, thus preventing timely management interventions. The automation and precision of real-time progress monitoring are critical for the concrete pouring process, significantly impacting both the efficiency and overall quality of the project. To address the common detection challenges like fuzzy boundaries and partial occlusion, an improved semantic segmentation-based framework was proposed for real-time progress tracking. The ECA mechanism and ASPP module were integrated into the U-Net to achieve precise segmentation. Subsequently, the actual progress was evaluated, and a work state identification algorithm was utilized to detect productivity-related issues, such as delays and interruptions. The approach was validated on a real construction project, achieving 90.85% precision in segmentation. This study provides theoretical references for real-time progress monitoring and presents a feasible approach for construction productivity management.