Urban traffic flow estimation with noise measurements using log-linear regression

This study proposes the determination of a log-linear regression model for estimating average traffic flow rates using a single measured noise indicator. This model was trained and tested with noise and traffic count data collected over 400 days at a case study location in central Stockholm, Sweden. Through a comprehensive analysis of the correlation between various noise indicators and traffic counts, the best performing indicator was selected, enabling traffic flow estimation with an average day-wise RMSE of 2.31 vehicles per minute and percentage error of 7%. Different measurement campaign strategies were tested to assess their effectiveness in providing reliable training data, demonstrating that campaigns measuring over all hours of the day and all days of the week perform significantly better than campaigns restricted to typical weekday working hours. This study highlights the potential of noise-based traffic estimation as a complementary, cost-effective approach for enhancing real-time traffic monitoring and transportation assessment. • Regression model for estimating average traffic flow rate using noise measurements. • Model uses a single noise indicator most correlated with traffic counts. • Model testing performed with data measured over 400 days in central Stockholm. • Achieved average day-wise RMSE of 2.31 vehicles per minute and percentage error of 7%. • Application over urban noise sensor network can enhance real-time traffic monitoring.