| dc.description.abstract | Construction disruptions often cause schedule delays and budget overruns. Accurate disruption monitoring is crucial for the timely recovery of affected construction projects. This study proposes a computer vision-based (CVB) multiobject tracking (MOT) method for disruption monitoring in complex construction environments. This approach incorporates a sparse-optical-flow-based module for short-term undetected mask estimation and a deep re-identification (ReID) module for long-term occlusion handling. We also build a large-scale dataset containing 100 construction videos and 155,774 annotations to train the proposed MOT method. The experimental results show that our method outperforms state-of-the-art trackers across multiple representative evaluation metrics: the higher order tracking accuracy (HOTA), detection accuracy (DetA), association accuracy (AssA), localization accuracy (LocA), identification F1 score (IDF1), and identity switches (IDSW) are 61.6%, 57.9%, 66.4%, 91.1%, 64.0%, and 133, respectively. Additionally, field tests confirm the effectiveness of the MOT method in multiple truck tracking, arrival time recording, and disruption monitoring at construction sites. | |
