Research seminar "Traffic Sign Localization and Classification Methods: An Overview"

Abstract

In Advanced Driver Assistance Systems (ADAS), which are intended to assist the driver in the driving process, there is a need for the designing of a system for the detection, tracking and recognition of traffic signs in the vicinity of the vehicle. The goal is to maximize the safety of the passengers inside the vehicle and consequently all other traffic participants. Also, pipeline which consist of traffic sign detection, recognition and tracking has application in autonomous intelligent vehicles and in mapping and assessing the state of traffic infrastructure.

The aim of detecting traffic signs, within frames in video sequences, is to determine regions of interest where there is a potential candidate of traffic sign. Another point of view on traffic sign detection problem is as a segmentation problem where there are two classes: traffic sign and background class. Traffic sign recognition methods classify detected traffic sign candidates in specific predetermined classes or super-classes. To improve overall detection and recognition accuracy of traffic signs in the scene, traffic sign instances (real world signs) tracking  between adjacent frames in video sequences is conducted.

In this presentation current state-of-the-art methods for traffic sign localization and classification are presented. Also, in more detail, Integral Channel Feature detector and Convolutional neural networks are described. In the end, overall conclusion and future goals are provided.

CV

Ivan Filković studied Computer Science at the University of Zagreb, Faculty of Electrical Engineering and Computing where he received his Bachelor degree (BSc; 2011) and Master of Science degree  (MSc; 2013). While studying, he focused on the field of Computer Vision and because of that topics of his thesis are from this area. The topic of his Bachelor thesis was  "Object tracking based on adaptive search-window shift" and the subject of his Master thesis was "Evaluation of procedures for traffic signs recognition". The work on thesis was mentored by Professor Zoran Kalafatić, PhD. Presently Ivan is studying at FEEC on postgraduate doctoral study programme towards the Ph.D. academic degree in the scientific field of Computing. Currently he is employed on project VISTA which is a project whose overall objectives are strengthening of technology transfer and commercialization capacities of partner HEIs, transfer of existing computer vision applications from HEIs to SMEs, and developing new traffic- and transportation-related computer vision applications with commercial potential in collaboration with SMEs in the automotive industry sector. Specifically, Ivan is working on a system for the detection and recognition of traffic signs and its application in advanced driver assistance systems - ADAS. The goal of traffic sing detection in video sequences is accurate localization of the traffic sign and region of interest assessment where there is a potential traffic sign candidate. Procedures for traffic signs recognition assigns detected traffic signs to specific classes or categories.

Previously Ivan Filković worked in the Institute for Nuclear Technology (INETEC) on secondary verification system for positioning the manipulator based on Computer Vision methods.