Text based maximally stable extremal regions to detect vehicle plate location

  • Hendra Maulana Universitas Pembangunan Nasional “Veteran” Jawa Timur, Indonesia
  • Darlis Herumurti Institute of Sepuluh Nopember Surabaya, Indonesia
  • Anny Yuniarti Institute of Sepuluh Nopember Surabaya, Indonesia


The license plate recognition strongly support intelligent infrastructure systems, such as toll and parking payment application, toll monitoring application, traffic monitoring application, and so forth. Although it has shown promising performance, but some method may fail in a more complex situation, because of the complexity of such variation of the position and orientation of the plate, different illumination, different backgrounds, and objects of non-plate. For efficiency higher visual matching, some fast keypoint detectors and corresponding descriptions have been carried out in several research, such as FAST, SURF, BRISK, Harris Corner feature. In general, plate detection systems have two problems, namely where the plate is and how big is its size. In this paper, we present the number plate localization method based on text segmentation of unstructured standard plates. The algorithm is capable of detecting large number of candidate text regions and progressively removing those that tend not to contain text. The experimental results show that from 16 images with size 4208 x 3120 pixels with a complex background is 87.5% accuracy, and the average detection time is approximately 33.25 seconds. Based on the results, the MSER feature detector can find the text area well. It holds consistent color and high text contrast


Download data is not yet available.
Jul 13, 2018
How to Cite
MAULANA, Hendra; HERUMURTI, Darlis; YUNIARTI, Anny. Text based maximally stable extremal regions to detect vehicle plate location. Proceedings, [S.l.], v. 1, n. 1, p. 450-455, july 2018. Available at: <http://ojs.pnb.ac.id/index.php/Proceedings/article/view/936>. Date accessed: 17 aug. 2018.