Detection and classification of road signs for automatic inventory systems using computer vision

Abstract

This article describes an intelligent system that enables the automatic recognition of road signs from image sequences in road environments. The main difficulties the system has to deal with are related to changes in lighting conditions, obstacles blocking the view, the presence of objects with geometric and chromatic similarities and the absence of previous knowledge about their position and orientation. The application of different techniques allows the system to overcome this variety of problems. Therefore, the road sign recognition system is based on a first pre-processing of images, making use of information about road geometry to isolate those areas of the image where road signs may appear. The detection step uses colour and shape analysis to determine the regions of the image where potential road signs may be located. A third step focuses on recognition and classification using pattern matching and edge feature analysis. The proposed algorithm has been tested in different weather and lighting conditions and roads (one and two-lane roads and motorways), overall, a total of 1200 kilometres with a very high success rate of detection and classification as the experimental results show.

Artificial Bee Colony Optimization for Feature Selection of Traffic Sign Recognition

“…The existing approaches can be divided into three categories (Yin et al, 2015). In the first class, preprocessing methods are oriented to locate and recognise the traffic signs (Coronado et al, 2012) (Larsson & Felsberg, 2011). These methods require lots of preprocessing of the traffic images, and their generalisation capability and robustness are limited.…”

Fast Traffic Sign Recognition with a Rotation Invariant Binary Pattern Based Feature

“…Therein, L denotes the luminance, and a and b denote the color. Coronado [ 8 ] develops an intelligent system to achieve automatic traffic sign recognition in terms of dealing with the difficulties that arise from changes in lighting conditions and various obstacles. Hu [ 9 ] achieves the traffic sign detection based on the visual attention model.…”

Empirical Innovation of Computational Dual‐Loop Models for Identifying Vehicle Classifications against Varied Traffic Conditions

“…With this consideration, the capability of clarifying phases or states of traffic flows becomes a key in vehicle classification problems. Although tremendous studies have been reported on modeling traffic flow characteristic through multiple data sources (e.g., loops, video sensors, and floating vehicles) (Treiber and Kesting, 2011;van Lint and Hoogendoorn, 2010;Huang, 2010, 2012;Peláez Coronado et al, 2012), a challenge remains in identifying traffic phases by using the variables that could be directly calculated from the dual-loop data. Accordingly, two research problems inspired the authors' motivation to develop an empirical approach to innovate the length-based vehicle classification modeling and relevant computational algorithm against varied traffic flow phases at dual-loop detection stations.…”