A method for long-distance traffic load reconstruction by fusing multi-source information

Understanding the actual traffic load conditions on bridges is of significant importance for bridge safety and maintenance. Current traffic load reconstruction methods lack robustness in long-distance scenarios. To this end, a long-distance traffic load reconstruction method based on the fusion of multi-source data is proposed. Firstly, the coordinate attention mechanism is introduced to improve the YOLOX model. Next, the multiple vehicle sort (MultiVehiSORT) tracking algorithm, which incorporates the microscopic motion characteristics of vehicles, is proposed. Furthermore, a radar-vision fusion method is utilized to realize long-distance vehicle detection and tracking. Finally, spatiotemporal information and weight data of vehicles are fused, and traffic load characteristics are analyzed to further refine the weight information, thereby realizing complete reconstruction of traffic load information. Through the above, accurate reconstruction of traffic loads over longer distances can be achieved. This method is applied to a cross sea bridge - the Hong Kong-Zhuhai-Macao Bridge, and the results show that the mean average precision of improved YOLOX model realizes 97.71 %. The tracking multi-object tracking accuracy and the multi-object tracking precision of the MultiVehiSORT algorithm are 90.8 % and 76.6 %, respectively, with no ID switching phenomena occurred during tracking. This indicates that the method proposed in this paper realizes stable tracking of vehicle loads in long-distance. Based on this, the proposed multi-source information fusion method further realizes comprehensive reconstruction of long-distance traffic load information.