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

## TL;DR

This paper introduces a new method for improving traffic flow data quality using video frame rates and geospatial similarity.

## Contribution

A novel data repair model using LSTM and cross-sectional geospatial similarity for traffic flow data quality control.

## Key findings

- The proposed model outperforms in data repair under various sampling periods and missing rates.
- The method improves spatiotemporal similarity of traffic flow data collected via video.
- Experiments show competitiveness in traffic flow data quality control.

## Abstract

The quality of traffic flow data is very important to the effective management and operation of urban traffic system. At present, most traffic flow data used in traffic flow research come from road sensors, but the shortcomings of long sampling period and sparse sampling points affect the quality control of traffic flow data. To solve these problems, we propose a traffic flow data quality control method under video frame rate considering cross-sectional geospatial similarity. Under this framework, we design a video-based multi-section traffic flow data collection method to improve the availability of spatiotemporal similarity of traffic flow data. Further, combining the advantages of traffic flow data in space-time dimension under video frame rate, a data repair method based on cross-sectional geospatial similarity and piecewise interpolation is proposed, and a multi-sectional combined repair model based on LSTM is constructed. Experiments were carried out on several road cross-sections, and the results show that the proposed model has the best data repair effect under different sampling periods, different missing rates and different missing types, and has certain competitiveness in traffic flow data quality control.