DOI: 10.5176/2251-1652_ADPC12.10
Authors: Babak Vazifehkhah Ghaffari, Marzuki Bin Khalid, Mojgan Kouhnavard
Abstract:
The License Plate (LP) localization plays an important rule in License Plate recognition system for some countries which the LPs have varieties such as different shapes, Fonts and so on. The trouble revealed when the LP’s font is identically used on the body of cars or using handwritten fonts in LPs that need the more efficient system by high time consumption algorithms for character recognition process. Therefore, this paper presents a new method to solve LPs localization problems in Malaysia that the LPs’ shapes and fonts are obviously optionally selected by owners. In this method, first Histogram Equalization is applied for image improvement then the Otsu’s method has been exerted for global image thresholding. Consecutively, after Blob analysis applying to label objects, the best threshold boundaries are determined by modifying the initial threshold values which assigned to eliminate undesired objects. These values present the best condition of Otsu’s method due to Malaysian LPs localization system. At the end, the lowest and highest boundaries for heights and width of objects are determined and the LP is identified by new threshold’s boundary. This method performed in natural noise condition such as light reflection and also applied to LPs with different shapes and sizes. The results are achieved from exactly experimentation and the accuracy is 94{6e6090cdd558c53a8bc18225ef4499fead9160abd3419ad4f137e902b483c465} that is a valuable percentage because of the high varieties of LPs in Malaysia.
Updating...