UrbanLPR-Dataset: A Large-Scale License Plate Recognition Dataset for Travel Time Prediction

This repository contains the UrbanLPR Dataset, a large-scale dataset of license plate recognition data collected in Dongguan, China, designed to support research in urban traffic analysis and travel time prediction.

Paper

This dataset was created for our research paper, which has been accepted for publication in the journal Measurement.

• Title: Urban Road Network Travel Time Prediction Method Based on "Node-Link-Network'' Spatiotemporal Reconstruction: A License Plate Data-Driven WGCN-BiLSTM Model
• Authors: Weiwei Qi*, Bin Rao*, and Jiabing Wu (* co-first authors)
• Journal: Measurement (Accepted for publication)
• Corresponding Author: Jiabing Wu ([email protected])

Dataset Description

This dataset contains vehicle passage records from License Plate Recognition (LPR) cameras deployed at major intersections in Dongguan, China, from March 1, 2023, to March 20, 2023. All data has been fully anonymized to protect privacy.

The dataset is ideal for research in:

• Travel time prediction and estimation
• Spatiotemporal data mining and forecasting
• Graph-based traffic analysis
• Path reconstruction in sparsely sensored networks

File Structure

The dataset is provided as a .zip package containing the following structure:

UrbanLPR-Dataset_v1.0/
├── 2023-03-01.parquet
├── 2023-03-02.parquet
│   ...
├── 2023-03-20.parquet
├── distance.csv
├── intersection_map.jpg
└── vehicle_type_mapping.csv

Main Data Files (.parquet)

Each .parquet file contains the anonymized traffic data for a single day. The schema is as follows:

Column Name	Data Type	Description
vehicle_id	string	Anonymized 64-character unique vehicle identifier.
timestamp	datetime	The exact time a vehicle was detected.
intersection_id	integer	A unique ID for the intersection.
vehicle_type	integer	A numeric ID for the vehicle type.

Auxiliary Files

• distance.csv: A matrix containing the road network distance (in meters) between every pair of intersections.
• intersection_map.jpg: A map of the study area, labeling each intersection with its intersection_id.
• vehicle_type_mapping.csv: A table mapping the numeric vehicle_type ID to its Chinese and English names.

How to Cite

If you use this dataset in your research, please cite our paper.

@article{qi2025urban,
  title={Urban road network travel time prediction method based on “node-link-network” spatiotemporal reconstruction: A license plate data-driven WGCN-BiLSTM model},
  author={Qi, Weiwei and Rao, Bin and Wu, Jiabing},
  journal={Measurement},
  pages={118339},
  year={2025},
  publisher={Elsevier}
}