Title: Detecting text in license plates using a novel MSER-based method

Authors: Admi Mohamed; El Fkihi Sanaa; Rdouan Faizi

Addresses: IRDA Group, ADMIR Laboratory, Rabat IT Centre ENSIAS, Mohammed V University of Rabat, Rabat, Morocco ' IRDA Group, ADMIR Laboratory, Rabat IT Centre ENSIAS, Mohammed V University of Rabat, Rabat, Morocco ' IRDA Group, ADMIR Laboratory, Rabat IT Centre ENSIAS, Mohammed V University of Rabat, Rabat, Morocco

Abstract: A new license plate detection method is proposed in this paper. The proposed approach consists of three steps: the first step aims to delete some details in the input image by converting it to a grey-level image and inverse it (negative) and then use MSER for the extraction of text in candidate regions. The second step is based on a dynamic grouped DBSCAN algorithm for a fast classification of the connected region, and the outer tangent of circles intersections for filtering regions with the same orientations. Finally, a geometrical and statistical character filter is used to eliminate false detections in the third step. Experimental results show that our approach performs better and achieves a better detection than that proposed by Yin et al. (2014).

Keywords: text detection; MSER; circle overlapping; DBSCAN; license plate detection.

DOI: 10.1504/IJDATS.2020.111488

International Journal of Data Analysis Techniques and Strategies, 2020 Vol.12 No.4, pp.335 - 348

Received: 05 Nov 2018
Accepted: 07 Jul 2019
Published online: 30 Nov 2020 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article