Artificial Bee Colony Optimization for Feature Selection of Traffic Sign Recognition

Artificial Bee Colony Optimization for Feature Selection of Traffic Sign Recognition

Diogo L. da Silva (Polytechnic School of Pernambuco, University of Pernambuco (UPE), Recife, Brazil & Federal Institute of Pernambuco (IFPE), Palmares, Brazil), Leticia M. Seijas (Departamento de Computación, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina) and Carmelo J. A. Bastos-Filho (Polytechnic School of Pernambuco, University of Pernambuco (UPE), Recife, Brazil)
Copyright: © 2017 |Pages: 17
DOI: 10.4018/IJSIR.2017040104


This paper proposes the application of a swarm intelligence algorithm called Artificial Bee Colony (ABC) for the feature selection to feed a Random Forest (RF) classifier aiming to recognise Traffic Signs. In this paper, the authors define and assess several fitness functions for the feature selection stage. The idea is to minimise the correlation and maximise the entropy of a set of masks to be used for feature extraction results in a higher information gain and allows to reach recognition accuracies comparable with other state-of-art algorithms. The RF comprises as a committee based on decision trees, which allows handling large datasets and features with high performance, enabling a Traffic Sign Recognition (TSR) system oriented for real-time implementations. The German Traffic Sign Recognition Benchmark (GTSRB) was used for experiments, serving as a real basis for comparison of performance for the authors' proposal.
Article Preview


The traffic signs have considerable importance in the organisation of modern societies. In general, their visual properties are high delimited since they are designed to be easily identifiable by humans. Each government adopts laws and specific types of signs aiming to ensure the safety of each inhabitant and the correct movement of people and vehicles.

Traffic sign recognition (TSR) is an essential component of a driver assistance system (DAS). It enhances safety by informing the driver the speed limits or possible dangers such as icy roads, roads under maintenance, or crossing pedestrians. TSR algorithms find three main difficulties: 1) the poor image quality due to low resolution, bad weather conditions, and over or under illumination; 2) the rotation, occlusion, and deterioration of the signs; and 3) the limited memory and processing capacities in real-time applications such as advanced DAS (Zaklouta & Stanciulescu, 2012).

Furthermore, the traffic signs obtained like video and images have different appearance kinds. Including illumination, color and shadows at different days, seasons and weather, this becomes a problem to applications in real-time.

An efficient method for detecting signs and traffic signals is essential to allow real-time DAS. Given the recent research in Computer Vision and recent advances in technology, detection and recognition of traffic signals is already a reality. However, there are still significant challenges to be overcome, because this task still requires an enormous computational effort (Yin, Ouyang, Liu, Guo & Wei, 2015).

TSR are defined to cover two problems: traffic sign detection (TSD) and traffic sign classification (TSC). TSD is used for the accurate localization of traffic signs in the image space, while TSC handles the labelling of such detections into specific traffic sign types or subcategories (Mathias, Timofte & Benenson, 2013). Figure 1 depicts an example of a common street view with traffic signs marked with bounded boxes. Therefore, the TSD problem aims to define the bound boxes and the TSC classifies the image to a pre-defined type of signpost.

This paper proposes a hybrid system for real-time applications. The system has two stages. In the first one, we apply a swarm intelligence algorithm called Artificial Bee Colony (ABC) (Karaboga & Akay, 2008) for feature selection to be used in a TSR system. The second stage performs the classification of the traffic signal by using a Random Forest algorithm (Breiman, 2001).

Figure 1.

Example of a Common street view with traffic signs marked with bounded boxes (Ciresan, Meier, Masci & Schmidhuber, 2012)

According to Bastos-Filho and Guimarães (2015), Swarm intelligence algorithms have been successfully deployed to solve optimisation problems, and some of them have obtained interesting and promising results, mainly due to the exploitation abilities for fine-tuning in optimisation problems. Our proposal is an extension of a previous work (Silva et al., 2015), considering the definition and evaluation of several fitness functions. We also perform a parametrical analysis concerning the number of food sources in the ABC technique. The improvements achieved in this paper allows reaching recognition rates comparable with other state-of-art algorithms, but with a lower online execution time. We used the German Traffic Sign Recognition Benchmark (GTSRB) (Stallkamp, Schlipsing, Salmen & Igel, 2011) (Houben, Stallkamp, Salmen & Igel, 2013) for the experiments. This dataset is widely referenced in the literature (Mathias et al., 2013) (Ciresan, Meier, Masci & Schmidhuber, 2012), thus serving as a useful basis for comparison of performance for our proposal.

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 10: 4 Issues (2019): Forthcoming, Available for Pre-Order
Volume 9: 4 Issues (2018): 2 Released, 2 Forthcoming
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing