License plate recognition based on temporal redundancy

Abstract

Recognition of vehicle license plates is an important task that can be applied for a myriad of real scenarios. Most approaches focus on performing the ALPR steps using only a single frame for each vehicle. Hence, they might have their recognition rates reduced due to noise present in that particular frame. Therefore, in this work, we propose an approach to recognize vehicle license plates using temporal redundant information instead of selecting a single frame. We also propose two post-processing steps that can be used to improve the system accuracy by reconizing the vehicle appearance and querying a license plate database. Experiments demonstrate that it is possible to improve significantly the vehicle recognition rate using our proposed approaches on a dataset recorded at UFMG campus. Furthermore, this work also proposes a new benchmark composed of a dataset designed to focus specifically on the ALPR character segmentation step, a new evaluation measure and its evaluation protocol.