Traffic Sign Detection for Panoramic Images Using Convolution Neural Network Technique

Abstract

This research presents a method for panoramic traffic sign images detection for regulatory signs and guide signs especially blue and green sign. A new approach for detecting the signs inside a large panoramic image was considered. A convolution neural network technique was used as tools for detecting. In addition, the steps required are the technique used in conjunction with the convolution neural network technique by using Tensorflow training to improve the accuracy of traffic sign detection. Moreover, to improve the accuracy of image detection, some image processing technique was added. For example, adding brightness to an image. From the experimental results, detection of traffic signs from panoramic images (360°) by using trained convolution neural network model to improve a traffic sign detection, the accuracy from panoramic images (360°) is better than the traditional model.