Research on Real-Time Fusion and Intelligent Analysis Algorithm of Heterogeneous Sensing Data in Construction Site

Le résumé fourni par la source

In this paper, a fusion framework based on "edge-fog-cloud" three-level collaborative computing architecture is proposed to solve the problems of data island, lack of real-time, single analysis dimension and low intelligence level in current construction data management. In the realization of key technologies, a spatio-temporal dynamic alignment model based on BIM is constructed, and the rigid body transformation is used for spatial alignment, and the dynamic time warping (DTW) algorithm is introduced to realize time alignment. A multi-modal feature level fusion method based on graph neural network (GNN) is proposed to adaptively capture the complex dependencies between sensors. The design of intelligent analysis algorithms covers security early warning, progress detection, and resource scheduling optimization. The security early warning algorithm integrates deep learning and knowledge reasoning, performing anomaly detection through the Spatio-Temporal Autoencoder (ST-AE) model and introducing knowledge reasoning to enhance the interpretability and accuracy of early warnings. The progress detection algorithm integrates video, RFID and BIM point cloud data, and adopts the improved YOLOv7+DeepSORT framework and ICP algorithm to accurately evaluate the construction progress. The optimization of resource scheduling builds a model based on integer programming, aiming at minimizing the total cost in the construction period. The experimental results show that the framework performs well in spatial alignment accuracy, time synchronization performance, timeliness and accuracy of safety early warning, progress detection accuracy and system stability, which provides strong support for the development and application of smart site technology and can effectively improve the safety, efficiency and resource utilization of building construction sites.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.