Enhance of License Plate Matching Procedure Quality Using Augmented Probability Matrix Method

The issue of traffic data quality is an important part of the Smart City concept. Accurate and high quality traffic data serves cities not only for detailed traffic insights, online and offline traffic management, but also for future decision making on the im-plementation of strategic traffic measures. This paper focuses specifically on the processing and evaluation of data from constantly developing license plate recognition (LPR) systems and aims to develop a method that allows comparing text strings of license plate characters from different vehicle records and, without knowing the real (correct) shape of the license plate, determine whether the compared records belong to the same vehi-cle or not - in other words, whether they should be matched. The paper builds on previous research and a quality improve-ment method based on a probabilistic model for two measurement profiles. This paper focuses on the application, actualization and extension of the mentioned method to LPR systems with arbitrary number of measurement profiles. The proposed tool has been tested on real traffic survey data and is shown to be fully applicable for application to areas with arbitrary number of measurement profiles.