Abstract： Urban transportation systems are becoming increasingly complex. Traditional data processing methods are unable to handle massive, multi-source, and heterogeneous traffic data, leading to a prominent phenomenon of information silos and insufficient real-time analysis and decision-making capabilities. This paper proposes a processing framework consisting of data cleaning, feature extraction, and hierarchical fusion to address the challenges of multi-source and heterogeneous big data fusion in smart transportation systems. The aim is to enhance the consistency, usability, and value density of traffic data. An experimental environment encompassing multi-source data such as sensor data, video streams, and social media information is established. Meanwhile, the framework's practical effectiveness in data alignment, redundancy elimination, and real-time fusion is verified through a combination of stream computing and batch processing. Research confirms that this approach improves the accuracy of traffic state recognition and the timeliness of event detection, providing more reliable data support for applications such as dynamic route planning and congestion warning. This paper presents a feasible technical path for data integration and intelligent analysis in smart transportation systems, offering practical references for enhancing urban traffic management efficiency and promoting the intelligent transformation of transportation systems.
张明昭 Mingzhao Zhang (2025) studied this question.