Abstract
In this paper, a new low cost Arabic vehicle license plate recognition (LPR) method is proposed which is easily extendable to other plates. A novel LP segmentation technique with three sets of feature vectors was used with template matching to form the two main modules: license plate localization module and LPR module. This method was tested on more than 238 vehicle images taken from various scenes with different fonts and backgrounds from two Arab countries. The segmentation accuracy of the implemented system was 97.5% with a recognition accuracy of 99% for fairly distorted images. The presented model shows that despite the negative impact of shadows, cracks, dirt, and character separations, the system demonstrated an overall success rate of 92% for plate localization, 95% for plates segmentation, 92% for country and city recognition, and 99% for number segmentation and recognition. Combining all rates led to an overall system accuracy of 93%. Compared to many state-of-the-art LPR systems, this newly developed system uses 3 small training sets which cut the run times of the proposed solution to less than 5 seconds using the MATLAB R2008A running on a Compaq 8510W with 4G RAM. The results are comparable, and in some cases better with restricted conditions such as skew place, plate size, illumination and background.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call