Machine Learning Enabled Traffic Sign Detection System

Abstract

The Traffic Sign Detection system is a component of an advanced driver assist system that notifies and prompts the driver regarding traffic signals and boards in front. An well organized concurrent signal detection and warning structure are presented to assist better with the existing Intelligent Transport System (ITS) and to improve the safety systems for the identification of regulatory indicators. On-board cameras record real-time video and are associated with a computing device for further processing. The process includes image framing which is blurred and distorted with Gaussian noise because of the movement of the vehicle and ambient disturbances. Hence the input image is enhanced using the median filter and nonlinear Lucy-Richardson for deconvolution. This algorithm is best suited for implementation due to its efficiency in providing an optimal and effective graded output of the processed image. Colour segmentation is performed using Y CbCr colour spacing following shape filtering algorithms using template matching. Then, using processed colour-corrected samples, the required sign is extracted as colour and shape from processed photos, allowing the sign to be distinguished from its foreground and background. The role of the classification module is to find the category of noticed traffic indications captured utilizing Multilayer Perceptron neural systems. Compared to other available systems, the proposed system outshines in every aspect treated to obtain the optimum output. The proposed method is one of the major applications of machine learning which uses Lucy-Richardson and the colour segmenting process. The developed system is implemented efficiently and results close to proximity are obtained.