Automated License Plate Recognition of Philippine License Plates Myra Josephine S. Villanueva* Joel P. Ilao** Computer Technology Department College of Computer Studies De La Salle University – Manila myra_villanueva_3@yahoo.com* ilaoj@dlsu.edu.ph** Chloe Michelle K. Cervania Vincent Spencer Y. Ku Desmond Carvey T. Ragos Ty Computer Technology Department College of Computer Studies De La Salle University - Manila Abstract Automated License Plate Identifier (ALDEN) is a License Plate Recognition (LPR) System developed to recognize images of Philippine license plates. ALDEN captures images of an approaching vehicle when a sufficient distance from the camera is obtained, localizes the license plate from the scene and reads its content. This system is designed to be robust against changes in illuminations, and to a limited degree, correct perspective distortions in the acquired images, while still maintaining real-time performances. This paper describes the design and implementation of the ALDEN. Pre-processing techniques are first applied to the acquired raw images to correct uneven illumination, and perspective distortion. The license plate is then extracted from the visual scene, binarized, and segmented into characters using knowledge of the size and locations of characters in a license plate area using the Philippine License Plate format. Character Identification is then performed by segmenting the characters into sixteen regions, and comparing the regions with character templates using correlation as a similarity measure. Performance tests using this technique tested on 300 character images yield an average recognition rate of 93.75 percent. 1 Introduction License Plate Recognition (LPR) systems are used for automated vehicle management and identification, such as entrance admission in highways and establishments, security, parking control, and road traffic control. A number of commercial softwares have already been developed for this purpose[4][5]. However, these softwares are developed for a particular license plate format, and therefore, incapable of accomodating different styles and formats such as license plates used in the Philippines [3]. In Philippine commercial parking lots, for example, vehicle registration is still done manually: the parking attendant encodes the license plates of an incoming vehicle and gives the parking ticket to the driver. This method, however, is prone to error caused by human fatigue or carelessness. Automatic automobile management through LPR can give a more consistent performance, and savings through reduction of labor costs to the company. There is one LPR system that has already been developed for Philippine license plate recognition, in the College of Computer Studies, De La Salle University – Manila. Cue, et. al [1] used fuzzy techniques in thresholding, and statistical methods for the character recognition. The average accuracy rate of character recognition for the their study is 86.67%. Inaccuracies can be attributed to the system’s inability to recognize skewed images, images taken outside the recommended proximity range, and license plates not having the recommended Philippine license format. Additionally, these systems lack pre-processing techniques that would make the recognition invariant to changes in the environment, such as, illumination caused by weather changes. Thus the recognition rate may be further improved by improving the image quality such that images become more compatible to the recognition technique. A whole system that will handle the acquisition of the image of a vehicle with the desired license plate, addressing the aforementioned issues and limitations, was developed. In other countries, such as in Saudi Arabia and Korea, several studies and approaches are being done to develop an LPR that are used in several applications such as those mentioned above. These applications are suited to their own country’s format. There are also a number of commercial LPR systems available, such as SeeCar, Perceptrics, and Pearpoint.[4][5]