TRAFFIC SIGN DETECTION BASED ON BIOLOGICALLY VISUAL MECHANISM
Xiaoguang HU *”, Xinyan ZHU *, Deren LI *
* State key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan, China
Michael. hu.07@gmail.com
Commission III/3
KEY WORDS: Vision Sciences, Optical, Detection, Urban, Automation
ABSTRACT:
TSR (Traffic sign recognition) is an important problem in ITS (intelligent traffic system), which is being paid more and more
attention for realizing drivers assisting system and unmanned vehicle etc. TSR consists of two steps: detection and recognition, and
this paper describe a new traffic sign detection method. The design principle of the traffic sign is comply with the visual attention
mechanism of human, so we propose a method using visual attention mechanism to detect traffic sign ,which is reasonable. In our
method, the whole scene will firstly be analyzed by visual attention model to acquire the area where traffic signs might be placed.
And then, these candidate areas will be analyzed according to the shape characteristics of the traffic sign to detect traffic signs. In
traffic sign detection experiments, the result shows the proposed method is effectively and robust than other existing saliency
detection method.
1. INTRODUCTION
With the development of society and economy, the traffic
problems become more and more serious, which become one of
the bottle-neck of modern city. Traffic accidents are paid much
attention by the governments of the world. Under this condition,
ITS (Intelligent Traffic System) is in need. Traffic sign is
important information of the road traffic and its detection and
recognition has attracted much attention in last decade because
of its importance. It involves the technologies of pattern
recognition, digital image processing, artificial intelligence,
computer vision and so no. However, traffic sign recognition is
still an unsolved problem as it can be changed easily depending
on its relative location and angle of view against camera and
surrounding condition such as weather and daytime.
In recent years, traffic sign detection and recognition problems
have drawn the attentions of many researcher. Sebastian
Houben establish a new probabilistic measure for traffic sign
colour detection and propose a novel Hough-like algorithm for
detecting circular and triangular shapes!!. Kyung-in Min etc
proposed a method which can recognize about 4 directional
road signs in Region of Interest and, the experiment are based
on unmanned ground vehicle. The method using Coherence
Vector of Oriented Gradients features with neural network
classifier is promised by R. Rajesh etc and, they prove the
results based on the combination of other features can acquire
better recognition rates"). Siti Sarah Md Sallah etc propose a
road sign detection and recognition algorithm for an embedded
application, which use HSI color space to segment the road
signs color and the shape to classify road signs"). Besides, the
visual attention mechanism is introduced to detect traffic
sign, which mainly for prohibition sign as it is very
important for traffic safety’. Jiang Yanhua etc use the
algorithm that is composed of color segmentation, shape
detection and pictogram recognition to solve the problem.
In the first step Ridge Regression is used to obtain a
precise segmentation in RGB color space. Recognition
process include a novel feature extraction involves OTSU
method? Chen Zhixie etc propose the system divided into two
phases. In the detection and coarse classification phase, they
employ the Simple Vector Filter algorithm, Hough transform
and curve fitting approaches. In the refined classification phase,
the Pseudo-Zernike moments features and support vector
machines are used"). Lykele Hazelhoff and Ivo Creusen etc
report their works which different from others. They detect
present signs in street-level panoramic images and the signs also
need to be positioned besides the detection and classification!*!
Jung-Guk Park etc use machine learning algorithms to detect
traffic sign, and scale-space to handle the different scale of
traffic signs. In recognition phase, they introduce a novel
feature to distinguish different signs, which include the 64-
dimensional feature vector by the 4 Gabor filters and 16-
dimensional feature vector by the fast Fourier transform"). Liu
Yang etc describes an approach to using the location histogram
matching for the broken traffic signs recognition! ?),
The selective attention is the characteristics of human visual
system (HVS). We will unconsciously focus our attention at
saliency object and have not any prior knowledge when we saw
a natural scene. It is very important to find saliency area
because we can distribute finite computing resources according
to saliency object to reduce time consuming. The saliency
theory are used in many computer vision applications such as
image segmentation, target recognition and image search etc.
Visual saliency are closely related to the human experience and
cognition and many image attributes can cause people's
unconscious attention like contour, colour, edge and intensity
etc. Because the potential of saliency research in computer
vision, it is now investigated by multiple disciplines like
cognitive psychology, neurobiology, CV. The design principle
of the traffic sign is to attract the drivers which take into
account the visual saliency of human, e.g. they normally have
vivid colour, normalized shape etc. So, the use of the visual
attention mechanism is reasonable in traffic signs detection. In
this paper, we introduce a detecting method of the traffic sign
based on visual attention mechanism, and our goal is to detect
the prohibition signs because its importance in traffic safety. In
— 09S — 7 0€) aa]
oy
— mn Pa i MUS A-——— M OPER EY
