Traffic Sign Detection and 3D Localization via Deep Convolutional Neural Networks and Stereo Vision

Traffic sign detection is considered as a fundamental task for autonomous driving and Advanced Driver Assistance Systems (ADAS). Therefore, an algorithm to spacially locate the detected traffic signs is required in order to make these detections useful. In this paper, an algorithm to detect, classify and spatially locate multiple traffic signs in different scenarios is presented. Detection and classification are made simultaneously via YOLOv3, using RGB images. 3D sign localization is achieved by estimating the distance from the traffic sign to the vehicle, by looking at detector bounding boxes and the disparity map generated by stereo vision. Moreover, a new traffic sign dataset called LSITSD is created to solve the issues that most of traffic sign available datasets have, such as missing or incorrect labels, or the small amount of images provided. The obtained results show the performance and the robustness of the proposed algorithm in detecting the traffic signs and the accuracy of the 3D localization estimation of the detected traffic signs.