INTEGRATED SYSTEM FOR ROBOT CONTROL BASED ON TRAFFIC

SIGN RECOGNITION
MR. SATISH
M.Tech in VLSI design and embedded systems, Department of ECE,
Don Bosco Institute of Technology, Bangalore, Karnataka, India.
Suryawanshi.satish975@gmaill.com

MR. NAGARAJ.KYASA
Assistant. Proffesor,Department of ECE,
Don Bosco Institute of Technology, Bangalore, Karnataka, India.
Nagaraj.kyasa@gmail.com

ABSTRACT: This paper introduces a fast Traffic Sign Recognition system developed for a robot
control, participant in the Autonomous Driving Competition in the vishamay-2015 dbit Festival of
Robotics. The Autonomous Driving Robot performs detection and classification of traffic signs
based on the analysis of images acquired by a camera mounted on its chassis.The proposed
algorithm is composed of three processing stages: detection, pictogram extraction and
classification. After the two firsts processing stages, a binary pattern matrix is obtained by color
segmentation. In the classification stage two different neural networks were trained to recognize
the traffic signs . Experimental results show that the system precision is very close to 100%
whereas recall presents values above 90% in most of the signs. The proposed system also proves
to be reliable and suitable for real-time processing.

KEYWORDS: traffic sign recognition, neural network, autonomous driving robot. Intel atom
processor.

INTRODUCATION

Autonomous driving robots are known as a driverless car, autonomous car is an automated vehicle
is capable of fulfilling moving or traveling and has a control of robot without human instruction or
input[1]. It has the capacity to sense all its environmental obstacle and traffic sign. The robotic
cars are exits mostly as prototypes and experimental systems, the robotic field becoming more
popular because it attract the student and robotic gaming the award winning at the competition
which gives the student sense and visibility[2], some of the advanced robotic car which are
capable of detection and recognition of the traffic signs and warning signs, by design and
development of the robotic car which have fewer traffic collisions and reduction in theft due to
the robots self-awareness.the integrate system for robot control can be design by using Intel atom
processor ,two webcam one for obstacle avoidance other one for traffic sign recognition the
steering of the robot is controlled by a servomotor and for wheel the brushless DC motor is used in
this project the Intel atom processor is main controller[3]. the main aim of this project is to design
and development of real time traffic sign recognition (TSR) system for robot control which is
participating in the vishamy-2015festival of dbit, system should system to be capable to recognize
different traffic sign that appear on the road side of the track the traffic sign are represented with
different features like color.
Shape and pictogram these are the basic fundamental elements used for detection and recognition
steps in TSR control system usually the detection first step is based on the image sign color and
shape information of the image sign[4], after basic filtering of the traffic sign some of the most
frequently used color based segmentation are performed on the images. the thresholding is simple
type of image segmentation from a grayscale color image ,thresholding can be used to create
binary images. in this method of replacing each pixel in an image. The number of methods have
been proposed based on the large range such as RGB,HSV,RGBTOGRAY are used for sign shape
are extracted from edge detection using hough shape detector contour are found from fourier
descriptors finally recognition steps is to decided by the traffic sign and identify it and
classification can be done using template matching techniques like ANNs by this method 90% of
accuracy is achieved.

RELATED WORKS

As mentioned in the paper by V. Prisacariu, R. Timofte, K. Zimmermann, I. Reid, and L. Van

Gool The past year become computer technologies to a more advanced state. It is the time when
the research results started to affect virtually everybodys daily life, rather than being confined to
industrial production lines or other nice applications. this paper we mainly concerned on an
important aspect of the algorithms that have bring those uses within reach, namely suitable data
identification for computer vision.The main work that will get direct benefit for classification,
detection, segmentation, and image enhancement. Performance, efficiency, and good trade-offs
between those are the recurrent goals of our scholar.

As mentioned in the paper by S. Houben, In this paper, will gave a methodology studies the
sampling of individual units of the traffic sign detection literature, study of detection systems for
traffic sign recognition (TSR) for driver assist. We deferentially explained the contributions of
recent works to the different stages inherent in traffic sign detection: segmentation, feature
extraction, and final sign detection. While TSR is a well-startup research area, we explain open
research issues in the literature, including a dearth of use of publicly available image databases.

As mentioned in the paper by j. Stallkamp, M. Schlipsing, J. Salmen and C. Igel, In This paper
represents t he desi gn, devel opment and verification of an application to recognize
regulatory traffic signs on the road installed on Indian roads. The application is conceived as a
modul e of a dri ver assist ance syst em under devel opment, and an automat ed vehi cl e
adapted t o the local infrastructure. The application uses SVM which are trained and verified
with official synthetic images provided by the National Ministry of Transport.

As mentioned in the paper by j. Greenhalgh and M. Mirmehdi: in the project a novel system for
the automatic detection and recognition of traffic signs. The proposed system detects traffic sign
regions as maximum stable extremer regions (MSERs), which offers robustness to variations in
lighting conditions. Recognition is based on a serial of support vector machine (SVM) classifiers
that were trained using histogram of oriented gradient (HOG) features. The training data are
information is created from synthetic template traffic images that are freely available from an
online database thus, real footage road signs are not required as training data. The proposed system
is accurate at high vehicle speeds, operates under a range of real time processing, runs at an
average speed of 20 frame per seconds.

503
METHODOLOGY

Traffic Signs Algorithm: The proposed TSR system has three stages of processing
1. Detection
2. Pictogram extraction
3. Classification

DETECTION

In the detection frame is a captured from the webcam and some of the basic image processing
operation are carried out first RGB image is converted to the HSV in this conversion the
information obtained is brightness information making the detection of the particular color in
different ranges, the traffic sign is detected by thresholding the image represent in the HSV range.
The HSV ranges indicates the red and blue color in the signs in the opencv the limit of H is 0-180,
The block of the proposed TSR system is shown in fig.1the output of the HSV is thresholding
where separate out regions of images corresponding to objects.

Frame acquisition Pictogram

Detection Extraction

RGB to HSV RGB to Grayscale

Thresholding Thresholding

Region size
Region size
filtering
filtering
Classification

Logic OR

Neural Network

Fig.1.overview of the proposed TSR system

PICTOGRAM EXTRACTION

This the second stage after the image detection in this pictogram extraction stage inner traffic sign
is segmented, extracted and combined with the external contour which is obtained by the detected
image and finally this will create the full binary image pattern for classification. Again here the
RGB image is converted to grayscale image for better look some of the filtering operation is done

504
for better visibility. the logical operation is done with external contour and inner sign for binary
pattern .

CLASSIFICATION

The recognition stage is classification in which open source library is used here feed- forward
neural network library as a classifier to recognize the sign, traffic sign NN has three layer first
input layer second hidden layer and final is output layer with different neuron, the output neuron
layer has equal number input traffic sign The neurons from the hidden and output layers use the
logistic regression activation function. Logistic regression outputs a value between 0 and 1,
indicating a likeliness of the input pattern belong to one of the class signs. The NN assigns each
binary input pattern the type equal to with the output neuron that has the highest probability.

EXPERIMENTAL RESULTS

Detection: This system uses digital camera which mounted on the robot as the robot travels the
webcam will capture frames of the environment around the road side and on the track, the traffic
sign is detected is shown in fig2.the RGB color image is captured is converted to HSV color space
this HSV represents the chromatic color information percentage will tells the range of the RGB
color. by applying the thresholding external counter is obtained by separating the inner sign and
red color, the triangle box is shown in fig.2 to obtain the external counter maximum and minimum
vale are checked for better region size filtering.

Fig.2. traffic sign is detected and detected external contour

Pictogram extraction: once the traffic sign is the image is obtained inside symbol is extracted
original RGB frame. This sub-image is scaled/normalized to 50x50pixels and converted to
greyscale Two thresholding are applied to the sub-image in order to segment the interior

505
pictogram: one for dark pictograms and another for white pictograms. For a specific traffic sign,
both thresholding are always applied, and while one extracts the pictogram correctly the other
yields nothing, depending on the interior pictogram color to be either white or black. As shown in
fig3.

Fig.3. result of pictogram extraction

CONCLUSIONS

In this project robot is controlled based on traffic sign in real time application ,intel atom is main
processor .openCV is library is used for real time image processing, neural network is for traffic
sign recognition.

FUTURE DEVELOPMENT

This project can be further improved for the low power design. It can also be further improved by
implementing using high resolavation webcam and advanced sensors.

REFERENCES

T. Moura, J. Teixeira, F. Tuna, F. Moreira, A. Valente, V. Filipe, S. Soares. Reconhecimento de

Sinais de Trnsito para Prova de Robtica de Conduo Autnoma. Proceedings of 19th
Annual Seminar on Automation, Industrial Electronics and Instrumentation (SAAEI12),
pp.645-649, 2012
X. Qingsong, S. Juan, and L. Tiantian, A detection and recognition method for prohibition
traffic signs, in Proc. Int. Conf. IASP, pp.583586, April 2010

506
E. Krsak, S. Toth, Traffic sign recognition and localization for databases of traffic signs, in
Acta Electrotechnica et Informatica, vol.11, no. 4, pp. 31-35, 2011.
S. Houben, A single target voting scheme for traffic sign detection, in Proc. IEEE IV Symp., pp.
124129, June 2011
F. Ren, J. Huang, R. Jiang, and R. Klette, General traffic sign recog- nition by feature
matching, in Proc. 24th Int. Conf. IVCNZ, pp. 409414, November 2009
S. Xu, Robust traffic sign shape recognition using geometric matching, IET Intell. Transp.
Syst., vol. 3, no. 1, pp. 1018, March 2009

507