Multi-Pole Road Sign Detection Based on Faster Region-based Convolutional Neural Network (Faster R-CNN)

Building an approach system that is able to serve various types of traffic signs is a challenge. The important stages in handling an object are finding objects, dividing them into several categories, and marking objects with bounding boxes. However, in reality, monitoring traffic signs objects is quite difficult because it is based on various factors such as; other closed objects, driving times, or traffic sign conditions. This study aims to measure the level of precision in monitoring traffic signs (detection speed of 4–6 frames per second) from video recording (single camera) using the Faster Region based Convolutional Neural Network (Faster R-CNN) algorithm. The traffic sign detection system uses the Faster R-CNN algorithm with Inception v2 model which is implemented in the TensorFlow API framework. The Faster R-CNN consists of 2 different modules. The first module is a deep convolutional neural network which functions to build the area to be detected, which is called the Regional Proposal Network (RPN), and the second module is the Fast R-CNN detector which functions to use the previously proposed area. This system is one unit, a detection network based on the results of the manufacture and testing of a traffic sign detection system based on the Faster R-CNN method, so it can be shown that there is no difference in the results of detection of traffic signs in day and night conditions. Where the precision testing for detection of traffic signs during the day and at night is 100%.