Issue Number: Vol. 3, No. 3
Year of Publication: 2013
Page Numbers: 54-69
Authors: Charbel Fares
Journal Name: International Journal of New Computer Architectures and their Applications (IJNCAA)
- Hong Kong


In this paper we are presenting an intelligent system for License plate recognition. We need to specify the shapes that need to be detected as well as its dimensions. A comparative study is done in order to show the most effective contour detection method for this application. The masks used for this comparison are: Sobel, Canny, Prewitt, Roberts, Log and Zerocorss. The study shows that Sobel is the most effective for such type of applications. Then three methods were used for the object extraction. Two known methods were used: Hough Transform and Watershed. The third method is a new Hybrid Segmentation algorithm. Results show that the proposed algorithm is the fastest with a time of execution of 0.8 seconds for images of size 658 X 486 pixels. The recognized license plate is an input to an automatic number extraction. This phase is decomposed into three steps. The first step is the cleaning phase. It is then followed by the segmentation phase. And finally the number recognition step is applied.