Traffic flow data quality control under video frame rate considering section-level geospatial similarity.

Chen, Y; Lu, J · 2025 · PubMed Central

DOI: 10.1371/journal.pone.0320567

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical issue of traffic flow data quality control, which is foundational for Intelligent Transportation Systems (ITS) applications such as congestion warning and signal control. The authors identify two primary limitations in existing research: traditional road sensors suffer from long sampling periods (often 5–15 minutes) and sparse spatial coverage, leading to data that lacks the high-resolution spatiotemporal similarity required for accurate analysis. To overcome these challenges, the study proposes a novel framework that leverages video-based data collection at video frame rates to capture multi-section traffic flow data, thereby enhancing both temporal density and spatial correlation. The methodology consists of three integrated components. First, a video-based multi-section traffic flow data collection method is designed to gather high-frequency data across multiple road cross-sections, addressing the sparsity issues inherent in single-point sensor data. Second, a preprocessing and repair method is developed based on cross-sectional geospatial similarity. This approach utilizes piecewise interpolation to handle continuous and random missing data, leveraging the strong spatial correlations between adjacent sections. Third, a multi-section combined repair model based on Long Short-Term Memory (LSTM) networks is constructed. This deep learning model is designed to capture the complex spatial dependencies across multiple sections and the temporal dynamics of traffic flow, providing a robust mechanism for data imputation. Experiments were conducted on several road cross-sections using data from Nanjing Transport. The results demonstrate that the proposed LSTM-based multi-section model achieves superior data repair performance compared to existing methods. Specifically, the model exhibited the best repair effects across various conditions, including different sampling periods, missing rates, and missing data types. The findings indicate that integrating video-frame-rate data with cross-sectional geospatial similarity significantly improves the accuracy and reliability of traffic flow data reconstruction. The significance of this work lies in its contribution to more precise traffic monitoring and management. By shifting from sparse, low-frequency sensor data to high-frequency, multi-section video data, the study provides a more comprehensive grasp of traffic states, including flow, speed, and occupancy. The proposed framework offers a competitive solution for data quality control, enabling more reliable inputs for downstream ITS applications. This approach enhances the ability to manage urban traffic problems such as congestion and pollution by ensuring that the underlying data accurately reflects real-time traffic conditions with high spatiotemporal fidelity.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• traffic density