A Framework for LED Signboard Recognition for the Autonomous Vehicle Management System

Abstract

An electronic road sign is an important instrument for providing real-time traffic-related information in the field of intelligent vehicle management systems. In most cases, electronic signboards present a piece of complex text information in which each character is made up of a matrix of light-emitting diodes lamps, referred to as LED text. LED dot matrix displays are also widely used to display notifications and content in a variety of applications. A matrix with a defined number of rows and columns is used to represent a single character. Since it demonstrates discontinuity, the LED text is difficult to detect. To do so, we have proposed a digital image processing-based recognition technique in this paper. Between classes, the variance technique is implemented for converting gray-scale images to binary images. We have used the Sobel masking operator to detect the cell region from the LED text. An improved optical character recognition (OCR) technique is then applied to the normalized LED text images for recognition purposes. The key contribution of this paper is to detect and recognize discontinuous LED texts from different environmental conditions that can be used to assist driver-less vehicle management systems. Our proposed framework has achieved a recognition rate of 84.4%.