BIM-integrated motion tracking for structural construction behaviour analysis using a weighted Kalman filter

Sustainable green construction requires intelligent monitoring systems to minimise resource wastage and improve operational efficiency. However, existing approaches lack effective motion trajectory tracking (MTT) for accurate behaviour analysis in dynamic construction environments. To address this issue, a building information modelling (BIM) integrated MTT and activity classification framework using a novel weighted Kalman filter is proposed. Initially, construction site videos were processed through frame conversion and enhanced using pre-processing based on Chapman contrast limited adaptive histogram equalisation. Objects were detected using S-YoloV8 and tracked under occlusion using Samurai (segment anything model-based unified and robust zero-shot tracker with motion-aware instance-level memory). Motion trajectories and behavioural features were extracted and classified using the proposed DCD-GeluCNN. The integration with BIM enabled computation of a waste score index for sustainability evaluation. The experimental results were satisfactory, attaining a classification accuracy of 99.03% under the used experimental settings, and with 1.8–13% better performance compared with baseline models. Although the results show that the proposed framework is promising, further validation on large-scale and more diverse real datasets is needed. The framework enables real-time monitoring, optimises resource utilisation and facilitates data-informed decision making for sustainable green building.