Accuracy of a Traffic Noise Model Using Data from Machine Vision Technology

Abstract

The FHWA Traffic Noise Model is used in a variety of applications to predict noise levels at locations of interest in the vicinity of streets and highways. The model's traffic noise source characterization process requires the user to supply the volumes and speeds for each vehicle class for each lane of roadway to be modeled. Using these parameters, along with algorithms developed from typical vehicle pass-by data, the program calculates the vehicle noise emission levels for each vehicle class. The accuracy of the user-supplied traffic parameters, therefore, affects the final noise level prediction. Automated traffic data collection technologies offer an attractive alternative to manual data collection for high-volume roadways. The accuracy of noise model predictions was investigated for the machine vision technology, which uses video-based traffic detection with a procedure developed to determine the effect of video equipment setup parameters on traffic data acquisition accuracy, and the range of expected errors was found to be ± 12%. It was found that the range of traffic volume errors produced a range in predicted noise level error of less than ±0.6 dB and that the range of traffic speed errors produced a range in predicted noise level error of less than ±1.0 dB. The results of the study provide a procedure to help machine vision users establish camera position parameters with specific reference points to limit noise prediction errors to acceptable values.