An intelligent approach for robust detection and recognition of multiple color and font styles automobiles license plates: A feature-based algorithm

• Published in: 2014 International Conference on Audio, Language and Image Processing pp. 956 - 961
• Main Authors: Bhutta, M. Usman Maqbool, Mahmood, Hasan, Malik, Hafiz
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2014
• Subjects: Correlation; Feature extraction; Image color analysis; Image segmentation; Licenses; Neural networks; Vehicles
• ISBN: 9781479939022; 1479939021
• DOI: 10.1109/ICALIP.2014.7009936

This paper presents a robust real time feature based algorithm for localization and recognition of multiple color and font style license plates. The algorithm uses digitized license plate images for robust detection and recognition. The resolution of the captured image is adjusted and the brightness and contrast are set to appropriate levels. An optimal threshold is applied to detect white regions of the enhanced image. The algorithm uses segmentation and component labeling to create candidate regions intelligently. For localization of the license plate, features of each region are extracted and analyzed by rectangularity percentage of the automobile's license plate. The use of color filter renders the algorithm more robust on license plate localization. Moreover, we use HSV color space to find the specific color regions. After localization of license plate in small sized image, the proposed algorithm uses rapid intelligent method for localization of license plate form the original image. The localized license plate image is cropped and processed by the recognition algorithm. Each character's region is identified by using segmentation algorithm based on the profile which includes character dilation and resizing for subsequent operations. In this paper, correlation optical character recognition and multi-layer perception neural networks techniques are used for recognizing the characters. The performance of the proposed algorithm is acceptable even in low-quality or bad light images.