Fine-grained progress tracking of prefabricated construction based on component segmentation

The progress management for prefabricated construction facilitates on -time project delivery. Though noninvasive methods have efficiently been applied to progress recognition of prefabricated construction, existing challenges limited the recognition accuracy, such as partial feature occlusion. To enable precise and efficient progress management, an instance segmentation -based framework was proposed for the automated progress tracking of prefabricated construction. HRFPN and the ECA mechanism were incorporated into the PointRend to achieve precise component segmentation. Drawing from the segmentation outcomes, the as -built progress of each construction layer was determined and compared with the as -planned progress to derive progress discrepancies. The method was implemented in a prefabricated construction project, achieving accurate component segmentation (mAP75 = 91.1%) and progress recognition. This study provides the theoretical references for precise progress tracking in prefabricated construction and offers practical insights for managing construction productivity, enabling future advancements in unmanned construction supervision.