Abstract

License plate detection (LPD) in context is a challenging problem due to its sensitivity to environmental factors. Moreover, the chosen font type in the license plate (LP) plays a vital role in the recognition phase in computer-based studies. This work is two folded. On one hand, we propose to employ Deep Learning technique (namely, You Only Look Once (YOLO)) in the LPD. On the other hand, we propose to evaluate font characteristics in the LP context.
This work uses 2 different datasets: UFPR-ALPR, and the newly created CENPARMI datasets. We propose a YOLO-based adaptive algorithm with tuned parameters to enhance its performance. In addition to report the recall ratio results, this work will conduct a detailed error analysis to provide some insights into the type of false positives. The proposed model achieved competitive recall ratio of 98.38% with a single YOLO network.
Some fonts are challenging for humans to read; however, other fonts are challenging for computer systems to recognize. Here, we present 2 sets of results for font evaluation: font anatomy results, and commercial products recognition results. For anatomy results, 2 fonts are considered: Mandatory, and Driver Gothic. Moreover, we evaluate the effect of the used fonts in context for the two datasets using 2 commercial products: OpenALPR and Plate Recognizer. The font anatomy results revealed some important confusion cases and some quality features of both fonts. The obtained results show that the Driver font has no severe confusion cases in contrast to the Mandatory font.