Content uploaded by Pranjal Protim Borah
Author content
All content in this area was uploaded by Pranjal Protim Borah on Dec 23, 2017
Content may be subject to copyright.
ISSN (e): 2250 – 3005 || Volume, 05 || Issue, 12 ||December – 2015 ||
International Journal of Computational Engineering Research (IJCER)
www.ijceronline.com Open Access Journal Page 24
A Study on Automatic Detection and Recognition Techniques For
Road Signs
1Purwi Saxena, 2Neha Gupta ,3Sabana Yasmin Laskar
4Pranjal Protim Borah
I. Introduction
The task of driving is completely based on visual information processing.
Road signs guide drivers for direction, gives current traffic situations, prohibit or permit certain directions and
warn them of any special road conditions. [7] The objective of this system is to detect and classify road signs
from within real-time colour images captured by an image-sensor on-board of the vehicle. The system attempts
to develop such a warning system which can alert the driver about the approaching road signs early enough to
prevent road accidents from happening.
But various challenges are faced while developing an Automatic Road sign recognition system, such as-
Appearance- The colour appearance of traffic signs is a crucial issue. As, the outdoor lighting
conditions varying from day to night may affect the colour display of road signs within images. Season
and weather conditions also affect the light strength. In addition, road sign images may be affected by
shadows from surroundings.
Road signs images also suffer from bluffing effect caused by the vibration of moving vehicle.
As the direction of sign is not always ideal, so the shape and pattern of sign within an image can be
affected. Sometimes the shape of signs also changes due to storms.
Road signs are often confused with other surrounding objects and many times, obstacles such as trees,
lamp posts, buildings, etc occlude the signs.[7]
These are unavoidable challenging issues faced during road sign detection and are the reason why this is
becoming a growing area of research since the 1980’s. [15]
The whole system is generally divided into three phases, namely-
Detection phase- In this phase, the road sign or a candidate area is searched within the obtained digital
image.
Classification/Recognition phase- Here, the candidate area or the detected shape is identified into a
road sign.
Finally, the obtained result is presented to the user.[12]
These phases are carried out using different techniques, which take advantage of the distinct colour and
shape based features of the road signs.
The techniques are divides into:
1.1. Colour Based Techniques
Colour is the most distinctive feature of a road sign. The applied colours are chosen in such a way that, they
are easily visible from far and they correspond to a specific wavelength in the visible spectrum. The most
commonly used colours are the primary colour (Red, Green, and Blue) and yellow (a secondary colour). Colour
Segmenting is the most commonly used technique, which extracts out the coloured road sign from the
surrounding. Arturo de la Escalera, [17] suggested the hue; saturation; intensity (HSI) system is very invariant to
ABSTRACT
Traffic signs provide valuable information about the road and play a vital role in safe and smooth
driving. But, while driving at high speed, these signs might get missed. To tackle with such problem,
the automatic road sign recognition system has been introduced. This system helps in improving
driver assistance system which is a development in Intelligent Autonomous Vehicles technique. This
paper discusses about the various existing methods used in detection and recognition of traffic signs,
the challenges that occur in dealing with live images and lastly concludes with a chronological and
brief tabulation of all the referred work.
Keywords: Road signs, Colour, Shape, Pixel, Detection, Recognition, Template.
A Study on Automatic Detection and Recognition…
www.ijceronline.com Open Access Journal Page 25
lighting changes. But, the disadvantage of HSI is that its formulas are nonlinear, and if special hardware is not
used, the computational cost is prohibitive. But, colour cannot be considered as a reliable feature for detection
[5]. Some widely used colour-based techniques are summarised below:
1.1.1. Colour Thresholding Segmentation. This technique isolates objects by converting grayscale images
into binary images. Image thresholding is most effective in images with high levels of contrast. The
advantage of obtaining first a binary image is that it reduces the complexity of the data and simplifies
the process of recognition and classification. Ghica et al. [1] used thresholding to segment pixels in a
digital image into object pixels and background pixels. The technique is based on calculating the
distance in RGB space between the pixel colour and a reference colour. [18], [19].
1.1.2. Region Growing. This approach uses a seed in a region as a starting point and expands as groups of
pixels with a certain colour similarity. The approach can be implemented in the HSI colour space. As it
requires a seed to start and ends when certain criteria are met, it may run into a problem when ending
conditions are not satisfied [6].
1.1.3. Dynamic Pixel Aggregation. This is another approach of colour segmentation. In this method it is
performed by introducing a dynamic threshold in the pixel aggregation process on HSV colour space.
The main advantage of dynamic threshold is to reduce hue instability in real scenes depending on
external brightness variation. This method has been implemented by Vitabile et al. [4].
1.1.4. Image Pre-Processing. The aim of pre-processing is an improvement of the image data that suppresses
unwanted distortions or enhances some image features important for further processing. This process
produces a corrected image that is as close as possible, both geometrically and radio metrically. The
main purpose of applying correction is to reduce the influence of errors or inconsistencies in image
brightness values [4].
1.1.5. RGB Transformation. A camera mounted on a moving car produces an RGB image. An important
part of colour-based detection system is colour space conversion, which means converting the RGB
image into another form that simplifies the detection process. This means separating the colour
information from the brightness information by converting the RGB colour space into another colour
space, which gives good detection abilities depending on the colour cue. This method has been used in
[4], [5] for detection purpose.
1.1.6. CIECAM97 Model. CIECAM97 is one of many colour appearance models. Shaposhnikovet al.[2]
used this model to perform pre-processing. It breaks down the illumination environment and the colour
stimulus. The CIECAM97 model uses the colour stimulus, the colour of white, the background, and the
surround field to calculate its representation of the stimulus colour. CIECAM97 represents the stimulus
colour using lightness, chroma and hue. In [3] the accuracy was found to be more than 90%.
1.2. Shape Based Techniques
Another important feature of road sign, as mentioned earlier is “Shape”. It can also be used for
detection purpose. It does not require colour information [7]. The system exploits a priori shape knowledge
about signs to select a candidate sign regions in the binary images obtained by the image sensor. Then a
classification is done on the shape using a similarity coefficient between a set of image samples representing
each road sign shape and a segmented region.[4],[19]
Due to lack of standard colours, Shape detection is preferred for road signs recognition as the colours found on
traffic signs changes according to illumination and it also reduces the search for a road sign regions from the
whole image to a small number of pixels. [12]
Some common approaches based on shape are:
1.2.1. Distance Transform Matching. In this technique, edges in the original image are found and then a
distance transform (DT) image is built. A DT image is an image in which each pixel represents the
distance to the nearest edge. Its advantage over other techniques is that it is capable of detecting objects
of arbitrary shapes when dealing with non-rigid objects. It also uses a template hierarchy to capture the
different shapes of object. Gavrila et al. [9] has used this technique for detection purpose. The method
is used to detect road signs both on-line and off-line with a detection rate of about 90%.
A Study on Automatic Detection and Recognition…
www.ijceronline.com Open Access Journal Page 26
1.2.2. Hierarchical Spatial Feature Matching (HSFM). Paclík et al.[11] developed a classification module
based on Hierarchical Spatial Feature Matching (HSFM) method. In the detection stage, a list of
regions where the signs are likely to exist was generated. This list is passed to the classification
module, by which each region is either labelled with the sign type found in this region, or marked as a
rejected region.
1.2.3. Regions Extraction. Since the images are taken outdoors, the obtained images are very noisy. Noise
reduction filters and morphological filters can be applied to enhance each region contour. A classical
region growing algorithm has been used to fix the region of interest coordinates. We do not consider
regions with an area less than 20x20 pixels since their information cannot be successfully recognized.
[4]
1.2.4. Similarity Detection. Vitabile et al. [4] performed Similarity Detection technique by computing a
similarity factor between a segmented region and set of binary image samples represent each road sign
shape. Both colour and shape information is considered for detection process. The performance of this
approach was reported to be good with triangular shape giving the lowest hit rate.
1.2.5. Template matching. Template matching is a technique that identifies the parts on an image that match
a predefined template [17]. It can be easily performed on grey images or edge images. Ohara et al.[8]
used template matching for sign recognition. A sub-area of size NxN is selected, and small defects and
noises are filled in or deleted. The pixel values are normalised by the maximum and minimum values
of the input sub-area. The size is also normalised depending on the template to be matched, and the
closeness with the template is calculated. A recognition rate of over 95% is achieved. [8]
1.2.6. Hough Transform. Hough transform is a method for estimating the parameters of a shape from its
boundary points. Gareth et al. [10] used Hough transform to isolate features of a particular shape within
an image. It is most commonly used for the detection off lines, circles or other parametric curves. It can
give robust detection under noise and partial occlusion. Its advantage is that it is conceptually simple,
easy to implement and handles missing and occluded data very gracefully. However Hough transform
is computationally complex for objects with many parameters and requires lot of memory that makes it
not a good choice for real-time applications.
1.3. Other Techniques
Some other techniques, apart from colour-based and shape-based methods are mentioned below-
1.3.1. Neural Network. Ohara et al.[8] used a small and simple neural network (NN) to detect the colour and
the shape of road signs. The original colour image is first treated by a Laplacian of Gaussian filter
(LOG). A colour NN classifier is then used to segment the image according to the colour under
recognition in RGB colour space. A shape NN is used after that to check whether each image contains
an object with the shape of a road sign. When a shape is found, template matching is applied for final
recognition. There are two distinct advantages of using neural networks. First, the input image does not
have to be transformed into another representation space. Second, the result depends only on the
correlation between the network weights and the network. Neural Networks are being implemented in
[4], [1], [15], [8] and etc.
1.3.2. Genetic Algorithm. Genetic Algorithm can be used to search for traffic sign in a scene image. Yuji et
al. [13] has used this approach. The image is matched by giving the gene information. Its advantages
are simple using, low memory demands, using of simple computation algorithm and ability of
parallelism. The disadvantage of genetic algorithm is non-deterministic work time and non-guarantee
finding of the best solution.
1.3.3. Laplace Kernel Classifier. Paclik et al. [11] used the Laplace kernel classifier to classify road signs.
The signs are divided into nine groups depending on their shapes and colours. This kernel is based on
Laplace probability density, and the smoothing parameters of Laplace kernel were optimised by the
pseudo-likelihood cross-validation method. The algorithm is tested on more than 4900 noisy images.
1.3.4. Nearest Neighbour Classification. Nearest Neighbour Classification is a straightforward and classic
type of classification. An image in the test set is recognised by assigning to it the label of most of the
closest points in the learning set. All images are then normalised to certain value. The image in the
learning set that best correlates with the test image is then the result. [16]
A Study on Automatic Detection and Recognition…
www.ijceronline.com Open Access Journal Page 27
II. Discussions
As has been mentioned earlier, the identification of road signs can be carried out by two main stages:
detection, and recognition. By invoking a combination of colour and shape, it is possible to take advantage of
both techniques to detect traffic and road signs. Each approach has its own positive properties and difficulties.
However, an adaptive hybrid approach can invoke one technique under certain circumstances and invoke the
other under different circumstances. Even when this adaptive approach is not in use, combining colour and
shape in any sign detection method has the advantage of using the information available from both sides of the
problem. Colour based detection is used over shape based, to increase the computational speed.
Techniques using shapes could be a good alternative when colours are missing or when it is hard to detect
colours. Shape-based techniques should be able to avoid difficulties related to invoking colours for sign
detection and robust to handle in-plane transformations such as translation, scaling and rotation.
Results of the template matching algorithms and the neural networks approaches are nearly the same.
However, template matching is a more time consuming process, because each time the candidate shall be
compared by each template, including its shifted versions.
On the other hand, training neural networks with lots of deformed versions of training set is possible.
Moreover, the processing time of the neural network does not directly depend on the number of the members of
training set. Thresholding uses only grey level value and no spatial information is considered. Therefore, the
major shortcoming of the threshold is that there is often an overlap between grey levels of the objects in the
breast and the background. Many techniques are robust and are able to detect and recognise the road signs with
high accuracy. But none of these can be totally immune to the problems faced by the automatic road sign
recognition system. Research in this field is still necessary and requires more attention.
III. Conclusion
The traffic sign recognition is a very helpful driver assistance technique for increasing traffic and driver
safety. The future intelligent vehicles would take some decisions about their speed, trajectory, etc. depending on
the signs detected. In this paper, a brief description and a review of the existing automatic road sign recognition
research has been given. It includes, description of the various methods, a short discussion on the techniques and
finally a tabulation of the different techniques is presented.
A Study on Automatic Detection and Recognition…
www.ijceronline.com Open Access Journal Page 28
References
[1] D. Ghica, S. Lu, and X. Yuan. Recognition of traffic signs by artificial neural network. presented at IEEE Inter. Conf.
Neural Networks, Perth, W.A., 1995.
[2] Shaposhnikov, Lubov N., Podladchikova, Alexander V., Golovan, Natalia, Shevtsova, A.,Hong, Kunbin, and Gao,
Xiaohong. Road Sign Recognition by Single Positioning of Space-Variant Sensor window. In Proc. of the J5th Int. Conf. on
Vision Interface, Canada, Calgary, 2002, pp. 213-217.
[3] X. Gao, N. Shevtsova, K.Hong, S.Batty, L. Podladchikova, A.Golovan,D. Shaposhnikov and V. Gusakova,. Vision Models
Based Identification of Traffic Signs. In Proc. Of the Is Europ. Conf. on Color in Graphics, Image and Vision, France, 2002, pp.
47-51.
[4] S. Vitabile, G. Pollaccia, G. Pilato, and F.Sorbello, Road signs recognition using a dynamic pixel aggregation technique in the
HSV colour space. In Proc. Int. Conf. on Image Analysis and Processing, Italy, 2001, pp. 572-577.
[5] Alberto Broggi, Pietro Cerri, Paolo Medici, Pier Paolo Porta. Real Time Road Signs Recognition. Proceedings of the 2007 IEEE
Intelligent Vehicles Symposium Istanbul, Turkey, June 13-15, 2007.
[6] A. Yuille, D. Snow, and M. Nitzberg. Sign finder, using color to detect, localize and indentify informational. Presented at
Sixth Inter. Conf. on Computer Vision, Bombay, India, 1998.
[7] Y.-Y. Nguwi and A.Z. Kouzani. A Study on Automatic Recognition of Road Signs. Presented at Cybernetics and
Intelligent Systems, 2006 IEEE Conference, Bangkok, June 07-09, 2006, pp.1-6.
[8] H. Ohara, I. Nishikawa, S. Miki, and N. Yabuki. Detection and recognition of road signs using simple layered neural network.
Presented at The 9th Inter. Conf. Neural Information Processing, Singapore, 2002.
[9] D. M. Gavrila. Traffic Sign Recognition Revisited. In Proceedings of the 21st DAGM Symposium fur Mustererkennung,
Springer Verlag, Bonn, Germany, 1999, pp. 86-93.
[10] Gareth Loy and Nick Barnes. Fast Shape-based Road Sign Detection for a Driver Assistance System. In Proc. of the IEEEIRSJ
Int. Conf. On Intelligent Robots and Systems, Vol. 1, 2004, pp. 70 - 75.
[11] P. Paclik and J. Novovicova. Road sign classification without color information. Presented at Sixth Annual Conf. of the
Advanced School for Computing and Imaging, Lommel, Belgium, 2000.
[12] Priyanka Satish Tekadpande, Ramnivas Giri. Recent Developments in Traffic Signs Recognition Techniques. In International
Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249 – 8958, Volume-1, Issue-5, June 2012.
[13] Yuji Aoyagi and Toshiyuki Asakura. A Study on Traffic Sign Recognition in Scene Image Using Genetic Algorithms and
Neural Networks. Published in Industrial Electronics, Control, and Instrumentation, 1996, Proceedings of the 1996 IEEE IECON
22nd International Conference, Taipei, August 05-10, 1996, pp.-1838-1843, vol.3.
[14] G. Jiang, Y. Zheng, and Y. Choi. Morphological skeleton analysis of traffic signs on road. Presented at IEEE Conf. Systems,
Man, and Cybernetics, Beijing, China, 1996.
[15] Arturo de la Escalera and Miguel Angel Salichs. Road Traffic Sign Detection and Classification. IEEE Transactions on Industrial
Electronics, Vol. 44, Issue 6, 1997, pp 848-859.
[16] S. Escalera and P. Radeva. Fast Grey scale Road Sign Model Matching and Recognition. Recent Advances in Artificial
Intelligence Research and Development, J. Vitria et.al. (Eds.) IOS Press, 2004, pp. 69-76.
[17] Yves Berube Lauziere, Denis Gingras and Frank P. Ferrie. A Model-Based Road Sign Identification System. In Proc. of the
IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, Vol.1,2001,ppI-1163- 1-1170.
[18] Wen-JiaKuo and Chien-Chung Lin. TWO-STAGE ROAD SIGN DETECTION AND RECOGNITION. Published in Multimedia
and Expo, 2007 IEEE International Conference, Beijing, July 02-05, 2007, pp. 1427-1430.
[19] Salem Saleh Al-amri, N.V. Kalyankar and Khamitkar S.D. Image Segmentation by Using Thershold Techniques. In Journal of
Computing, Volume 2, Issue 5, May 2010, ISSN 2151-9617.
