Automatic recognition of Ethiopian license plates

This paper presents an automatic recognition system for Ethiopian license plates. The proposed system has three major components: plate detection, character segmentation and character recognition. We used Gabor filters for plate detection, connected component analysis for character segmentation, and a correlation based template matching method for character recognition. In addition to the correlation value, the recognition process is supported by color analysis techniques and location information of characters. To test the performance of the system, a dataset of 350 car images of RGB color format were collected from moving cars under different angle, distance, motion and illumination conditions. Test results on the dataset showed plate detection rate of 88.9%, character segmentation accuracy of 83.9% among detected plates, and character recognition rate of 84.7% among correctly segmented characters. We have also achieved an overall plate recognition rate of 63.1%, and the whole recognition processes takes 2-5 seconds depending on whether post processing operations are needed or not.