Measuring License Plate Recognition Accuracy in Privacy-Preserving Image Analysis

Abstract

Today, information gathering is everywhere. For this reason, hiding information on personal data is increasingly becoming more important. For this study, a model that is developed by others in the research group has been quantitatively tested on a specific dataset. Images in this dataset consists of vehicles and license plates. This model removes features on license plates images that contribute to recognition. For this model, metrics such as correct license plate coordinates and value matching has been researched. It is shown that the performance degrades as levels of this privatization is increased.

Faculty Supervisor: Dr. Ivan V. Bajic, School of Engineering Science, Simon Fraser University