Automatic vehicle identification by plate recognition

In this thesis, a new algorithm for automatic vehicle identification has beenproposed. Identification is made by license plate recognition (LPR) algorithm.License plate recognition is a form of automatic vehicle identification and is analgorithm that identifies the vehicle by recognizing the license plate automatically. Inthis study, a simple but effective algorithm is presented for vehicle?s license platerecognition system.There are some other methods to identify the vehicles such as bar code-basedidentification systems, radio-frequency identification systems. In bar code-basedidentification and radio-frequency identification systems, the methods requireexternal components installed on vehicles for automated identification. On the otherhand, license plate recognition technology identifies the vehicle using only its licenseplate. Since every vehicle carries a unique license plate, no external cards, tags ortransmitters need to be recognizable. Due to this reason, license plate recognitionalgorithm is the most powerful and useful technique for automatic vehicleidentification.The proposed algorithm consists of three major parts: Extraction of plateregion, segmentation of plate characters and recognition of plate characters. Forextracting the plate region, edge detection algorithms and smearing algorithms areused. In segmentation part, smearing algorithms, filtering and some morphologicalalgorithms are used. And finally statistical based template matching is used forrecognition of plate characters.This algorithm operates on inactive real images and the system is designed for theidentification of Turkish license plates. The necessary codes for the proposed algorithm werewritten in Matlab software. To see the overall performance, the proposed algorithm has beentested over a large number of images. The images were taken on different time periods of theday and also these test images were taken under various illumination conditions. And it wasobtained that %97.6 success rate for the extraction of plate region, %96 success rate for thesegmentation of the characters and %98.8 accuracy rate for the recognition of plate characters,giving the overall system performance as %92.57 recognition rate.Moreover, some image processing techniques that have been used for license platerecognition were presented in this thesis. And the advantages and the use of these techniqueshave been explained.