Development of intelligent transportation system for Philippine license plate recognition

Abstract

Due to the growing need for surveillance and license plate identification in the country, a Philippine license plate recognition system was proposed by adopting accurate and suitable algorithms for each phase, namely, license plate preprocessing, plate localization, character segmentation and character recognition, in recognizing old and new license plates of both private and public vehicles and motorcycles. Preprocessing is based on improved Bernsen algorithm which is a shadow removal method. This algorithm performs local thresholding of a two-dimensional array image in which the grayscale input image is converted into a binary image. For plate localization, Connected Component Analysis is utilized for labelling regions and preserving the characters by removing the unwanted information. Prior to character segmentation, Hough transform is used for tilt correction of the extracted plate. From the corrected plate, the horizontal and vertical projections are obtained to locate the characters and the boundaries will be used as basis for segmentation. Character recognition involves two phases, namely, feature extraction using Dual-Tree Complex Wavelet Transform and classification using Artificial Neural Networks. The proposed algorithms are adjusted with parameters suitable for detecting and recognizing the format and specifications of Philippine vehicular license plates, resulting to practical plate character detection accuracy of 85.6667%, character recognition accuracy of 94.0183% and a practical license plate recognition accuracy of 72.83% (i.e. in the actual test run). However, the proposed system has some restrictions, which will be studied in the future. To make the techniques applicable in less restrictive working conditions, further efforts will be concentrated on the adaptivity of the set parameters on the working environment.