Vehicle Weight Estimation Based on Piezoelectric Sensors Used at Traffic Enforcement Cameras Experiences from the Norwegian system

In Norway, piezo-based traffic enforcement cameras (TEC) are used for speed limit enforcement, but it also has the technology to estimate vehicle weights. Weight data from such TECs can be useful for road authorities for purposes such as evaluating road wear and traffic safety. The first aim of this paper is to evaluate how accurate weight data from such systems are, while the second aim is to examine if one site's accuracy changes over time. We also want to see if calibration coefficients can be used to improve the accuracy of the acquired weight data. The last aim is to assess the weight data's accuracy classes and subsequent areas of application according to COST 323. The methods involve comparing dynamic weight data from a piezo-based TEC with corresponding static weights. This enables assessment of the system's accuracy and precision, and is performed on two dates to evaluate if it changes over time. Also, calibration coefficients are used to try to improve the accuracy. Lastly, accuracy classes are calculated according to COST 323. Our findings were that there was a systematic error between the dynamic and static weights, which increased with increasing vehicle weights. Also, the system's accuracy changed over time. One factor contributing to this was different mean static weights during the two data acquisitions, which was caused by different vehicle samples. Moreover, calibration coefficients generally improved the accuracy, but not the precision of the dynamic weight data. At their best, the dynamic weight data were found to have accuracy class B (10), which is accurate enough for pre-selection of vehicles and designing roads. The other weight data reached the less accurate accuracy class D (25).