International Journal of Innovations in Engineering and Technology (IJIET)
Accurate Number Plate Localization System
A.P.RamyaSri
Assistant Professor, Department of ECE
Saravana Kumar.G, Ram Kumar.T, Senthil Kumar.M, Manoj kumar.W
Final year, Department of ECE
SNS College of Engineering, Coimbatore
Abstract- Various number plate localization techniques have been developed in the past with variable results. In this
project, a localization method is implemented keeping in mind Indian vehicle number plate specifications. The beneficial
information from images is used in the determination of the number plates by utilizing image processing. The algorithm
makes use of morphological operations like reconstruction, dilation, erosion and exploits number plate edge
characteristics. The project makes use of color segmentation in the YCbCr color space for special number plates. For
initial analysis, the project underwent a color segmentation process with existing works. Skewed plates are corrected.
After conducting tests on a wide range of images with variable illumination and arbitrary camera angles, based on the
results achieved and finally the number plates are obtained. Now in the stage of analyzing and recording the number
plate by implementing that in MATLAB.
Key words – Color Segmentation, Edge Detection, Contrast Stretching and Localization.
I. INTRODUCTION
A vehicle registration plate is a metal or plastic plate attached to a motor vehicle or trailer for official identification
purposes. The registration identifier is a numeric or alphanumeric code that uniquely identifies the vehicle within the
issuing region's database.
¾
¾
¾
Vehicle license plates in India are of two types:
a) Public vehicles: Black color on a yellow background.
b) Private vehicles: Black color on white background.
Both these types exhibit characteristics high contrast, which can be used as a distinguishing factor in isolating the
license plate regions from the surroundings.
The proposed algorithm is implemented by cascading the operations in a sequential manner. The color segmentation
and contrast stretching is done in the preprocessing stage. Using certain characteristics of license plate edges,
candidate license plate regions can be determined. From these candidate regions, the statistical properties of the
corners and vertical edges are tested against a threshold. The corner detection is carried out using the Harris Corner
detection algorithm.
Segmentation is one of the basic techniques in computer vision. From the engineering viewpoint, visible light
consists of a small part of electromagnetic radiation from 380nm to 780nm in the wavelength domain. Therefore,
color segmentation becomes an important role in the area of quality control, image processing, pattern recognition,
and computer vision.
A color space is a method by which we can specify, create and visualize color. A color is a mathematical
representation of a set of colors. The most popular color models are
RGB (used in computer graphics);
YIQ, YUV, or YCbCr (used in video systems);
CMYK (used in color printing).
II. LITERATURE SURVEY
Kim, H.S. (1991) presented the method based on the Vertical edge using Hough Transform (HT) for extracting the
license plate. The proposed approach can be divided into four sections, which are, vertical edge detection, edge
statistical analysis, hierarchical-based license plate location, and morphology-based license plate extraction. Under
the experiment databases, which were taken from real scene, 9786 of 9825 images are successfully detected. The
average accuracy of locating vehicle license plate is 99.6% [9].
Lee, E.R. Kim, P.K. And Kim, H.J. (1994) used neural network for color extraction and template matching to
recognize characters. [11]. The former characterized by vertical gradient detection extracts candidate regions from an
Vol. 2 Issue 1 February 2013
200
ISSN: 2319 – 1058
International Journal of Innovations in Engineering and Technology (IJIET)
input image, while the latter conceptualized in terms of mathematical morphology aims to locate the license plate
fast and accurately[11].
Sirithinaphong, T. And Chamnongthai, K. (1998) describes that, some features of car license plate according to
motor vehicle regulation such as color information and shape are applied to determine the candidates of car license
plates. To certify the license plate, the characters, numbers and their patterns are recognized by back propagation
neural networks in windows which are opened in those boundaries of candidates. For the results of recognition by
neural networks, the candidate which has characters and number patterns according to motor vehicle regulation is
certified as license-plate region. Since the results of character-pattern recognition is used to certify the license plate,
the ability to license-plate extracting is enhanced and the car is identified simultaneously. The results of the
experiments with 70 samples of real car images show the performance of car license-plate extraction by 84.29%, and
the recognition rate is 80.81% [18].
Hongliang, B. And Changping, L. (2004) describes the method can improve the location rate only by the edge
statistics. The proposed approach can be divided into four sections, which are, vertical edge detection, edge
statistical analysis, hierarchical-based license plate location, and morphology-based license plate extraction. Under
the experiment databases, which were taken from real scene, 9786 of 9825 images are successfully detected. The
average accuracy of locating vehicle license plate is 99.6% [6].
Yang, F. And Ma, Z. (2005) describes the algorithm consists of two main modules: license plate region's rough
detection and license plate exact the location. The former characterized by vertical gradient detection extracts
candidate regions from an input image, while the latter conceptualized in terms of mathematical morphology aims to
locate the license plate fast and accurately. Experiments have been conducted for this algorithm, and 360 images
taken from various scenes were employed, including diverse angles, different lightening conditions and the dynamic
conditions. The algorithm can quickly and correctly detect the region of the license plate and the license plate
detecting the rate of success is 97.78% [22].
Babu and Nallaperumal (2008) developed A License Plate Localization using Morphology and Recognition. Sobel
mask was used to detect vertical edges in the input image. The resultant image was converted into a binary image.
Morphological dilation and erosion was applied then to extract the image. The Character Recognition algorithm
used the Cross Correlation Combined with Neural Network [1].
Shidore, Narote (2011) developed Number Plate Recognition for Indian Vehicles. Number plate extraction had been
done by using a Sobel filter, morphological operations and connected component analysis. Character segmentation
was done by using connected component and vertical projection analysis. Character recognition as carried out using
a Support Vector machine (SVM). Images had been taken in a different background, illumination conditions, and at
various distances from the camera to a vehicle. Images were resized to 1024*768. The accuracy of segmentation has
been reported in 80% and of recognition is 79.84 % [17].
III. NUMBER PLATE LOCALIZATION SYSTEM
The algorithm is implemented by cascading the operations in a sequential manner. The color segmentation and
contrast stretching is done in the preprocessing stage. Using certain characteristics of license plate edges, we
determine candidate license plate regions. From these candidate regions, the statistical properties of the corners and
vertical edges are tested against a threshold. The corner detection is carried out using the Harris corner detection
algorithm. Finally, the skew, was corrected. All structuring elements used for morphological operations are scaled
according to the size of the input image to ensure compatibility with images of different resolution.
The Algorithm is implemented by cascading the operations in a sequential manner. The color segmentation and
contrast stretching was done in the preprocessing stage. Using certain characteristics of license plate edges, the
candidate license plate regions can be determined. From these candidate regions, the statistical properties of the
corners and vertical edges are tested against a threshold. Finally, the skew, if any, was corrected. All structuring
elements used for morphological operations are scaled according to the size of the input image to ensure
compatibility with images of different resolution.
3.1 Color Segmentation:
In case of transport vehicles i.e. taxis and auto rickshaws, the license plate background is yellow with black
lettering. Sometimes, due to bad illumination, this results in low contrast and consequently failure in detection of
plate. It is difficult to accurately define different shades of yellow in the RGB color space; hence we convert the
image into luminance (Y) and chrominance (Cb and Cr) and replace the yellow region by using the following
thresholds. A region is classified as yellow if;
Vol. 2 Issue 1 February 2013
201
ISSN: 2319 – 1058
International Journal of Innovations in Engineering and Technology (IJIET)
Yellow
Fig.1.Replacing yellow regions by white
3.2 Contrast Stretching
Taking into account illumination variations, the image needs to have a well defined and crisp contrast. To do this,
we do contrast stitching with varying parameter values. This has its advantages over histogram equalization – the
resultant image does not become grainy if the illumination is low.
Depending on the need, fixed parameter values can be used if we have only images that are well lit. If we also have
poorly illuminated images especially those taken at night, these parameter values are computed dynamically by
taking the average of pixel intensities. The transformation relationship between f (x, y) and g (x, y) is:
Fig.2.Transform characteristics
3.3 Morphological Operations
The characters in the vehicle license plate can be considered as ‘regional minima’ or holes which are not connected
to the image border. An algorithm based on morphological reconstruction by erosion which can be used to fill these
holes is explained in:
G = Rf (fm)
Where, f is the original image,
G is the resultant image,
fm is the marker image.
Rf denotes reconstruction by erosion.
fm is set to the following values:
The original image (f) is then subtracted from the resultant image obtained after filling holes (G) to isolate the
regional minima. It is assumed that the road should intersect with one of the edges of the image. In practical
situations, this criterion is always met if the image is from a CCTV feed. Opening by a structuring element having a
vertical shape is done to further eliminate stray horizontal noise in the image. The subsequent processing steps rely
largely on the vertical information in the license plate; this does not affect the performance adversely. The formula
for image opening is:
A O B = (A .B) + B
Where A is the image and B is the structuring element.
Vol. 2 Issue 1 February 2013
202
ISSN: 2319 – 1058
International Journal of Innovations in Engineering and Technology (IJIET)
Fig.3.Result of subtracting the original image from filling image
3.4 Edge Processing
An efficient approach to localize license plate regions can be achieved using simple convolution and mathematical
operations. We first filter the image using the operator shown in fig. With the obtaining vertical edges:
1
0
-1
Fig.4.High Pass Mask
Vertical edges have been used in the past to localize license plates. But instead of taking absolute of the filtered
output, as is done in case of conventional edge detection, we separate its negative and positive parts to preserve
information about whether an edge is a dark-to-bright (white) or a bright-to-dark (black) edge. A unique
characteristic of license plate regions is the presence of a large density of black as well as white edges in close
proximity as shown in figure:
Fig.5.Number plate region edge
We then filter the images with a low pass filter. The neighborhood to be used in proximity criterion is defined by the
size of the low pass filter. The filtered edge images, shown in figure (b) and (c), are then multiplied to obtain the
result shown in figure (d). Only the regions with both black and white edges will result in a non-zero multiplication;
single edges will disappear. The merits of using the presence of both types of edges in a defined locality are evident
from the result: the license plate region is prominent, owing to the presence of both black and white edges. We then
remove areas smaller than a threshold and breaks that might occur due to large character spacing. This image then
acts as a mask to extract candidate regions from the original image.
Vol. 2 Issue 1 February 2013
203
ISSN: 2319 – 1058
International Journal of Innovations in Engineering and Technology (IJIET)
In the past, wavelet transform using an expanded Haar wavelet to identify candidate regions has been used. This
method takes the wavelet transform of the original gray image. They use the HL sub image, similar processing as
carried in our method can be carried out to obtain better results. Alternatively, variance can be calculated using a
sliding window to obtain similar results.
Fig.6.(a) Vertical Edges (b) White edge (c) Black edge
(d) Resultant
IV. CONCLUSION
Here's an accurate license plate which uses sequential operations to localize number plates is presented. The preprocessing stage provides accuracy in the recognition of number plates. By the process of edge detection and
morphological processing the edge characters are found clearly. The algorithm is quite reliable and it can adjust to
different lighting conditions due to the presence of the contrast stretching process. The output can also be stored in
the database for the further processing or checking of vehicle strategies.
In future, this algorithm is used to reject certain types of vehicle grills since they share all the characteristics of
license plates.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
Babu and Nallaperumal (2008) ‘A License Plate Localization using Morphology and Recognition’ IEEE, pp. 1-6,.
Bailey, D.G., Irecki D., Lim,B.K. and Yang,L.(2002) ‘Test bed for number plate recognition applications’, Proceedings of First IEEE
International Workshop on Electronic Design, Test and Applications ( DELTA’02 ), IEEE Computer Society.
Duan,J and Qiu,G.(2004) ‘Novel Histogram Processing for Colour Image Enhancement’ in: Proc. IEEE Int. Conf. Image Graph., pp. 5558.
Johnson, A.S., Bird, B.M.(1990) ‘Number-plateMatching for Automatic Vehicle Identification’ IEE Colloquium on Electronic Image and
Image Processing in Security and Forensic, April.
Kim, S.K., Kim, D.W. and Kim, H.J.(1996) ‘A Recognition of Vehicle License Plate Using a Genetic Algorithm Based Segmentation’
Proceedings of 3rd IEEE International Conference on Image Processing,V01.2., pp. 661-664.
Lotufo, R.A., Morgan, A.D., and Johnson, AS.(1990) ‘Automatic Number-Plate Recognition’ Proceedings of the IEE Colloquium on Image
analysis for Transport Applications, V01.035, pp.6/1-6/6, February 16.
Faradji, F. And Rezaie, A.H. (2007) ‘A Morphological-Based License Plate Location’ Proc., IEEE International Conference on Image
Processing, Vol.1, pp. I-57-I-60.
Hamidreza case, Mohammadreza Case and Alireza Case (2010) presented ‘New Morphology-Based Method for Robust Iranian Car Plate
Detection and Recognition’ International Journal of Computer Theory and Engineering, 2, No. 2,.
Hsieh, C.T. And Juan, Y.S. (2005) ‘Multiple license plate detection for complex background’ IEEE International Conference on Advanced
Information Networking and Applications, pp. 389-392.
Hontani, H. And Koga, T. (2001) ‘Character extraction method without prior knowledge on size and information’ Proceedings of the IEEE
International Vehicle Electronics Conference (IVEC’01), pp. 67-72.
Hongliang, B. And Changping, L. (2004) ‘A hybrid License Plate Extraction Method Based on Edge Statistics and Morphology’ Proc.,
17thInternational Conference on Pattern Recognition, Vol. 2, pp. 831-834.
Hung, K.M., Chuang, H.L. And Hsieh, C.T. (2007) ‘License Plate Detection Based on Expanded Haar Wavelet transform’ Proc., Fourth
International Conference on Fuzzy Systems and Knowledge Discovery, Vol. 4, pp. 415-419.
Jingbo Liu and Jin Liu and Weidong Jin (2010) ‘Robust Vehicle License Plate Localization Using TO-MACH Filter’.
Vol. 2 Issue 1 February 2013
204
ISSN: 2319 – 1058
International Journal of Innovations in Engineering and Technology (IJIET)
[14] Kim, H.S. (1991) ‘Recognition of a Car Number Plate by a Neural Network’ Proceedings of the Korea Information Science Society Fall
Conference, Vol. 18, pp. 259-262
[15] Kumar parasuram and Vasantha Kumar, P (2010) ‘An Efficient Method for Indian Vehicle License Plate Extraction and Character
Segmentation’ IEEE International Conference on Computational Intelligence and Computing Research.
[16] Lee,E.R., Kim, P.K. and Kim, H.J. (1994) ‘Automatic Recognition of a Car License Plate Using Color Image Processing’.
[17] Mohd Firdaus Zakaria (2010) ‘Malaysian Car Number Plate Detection System Based on Template Matching and Color Information’
(IJCSE) International Journal on Computer Science and Engineering, 02, pp. 1159-1164,.
[18] Park, S.H., Kim, K. I., Jung K. And Kim, H.J. (1999) ‘Locating car license plates using neural network’ IEE Electronics Letters, Vol. 35,
no. 17, pp. 1475-1477.
[19] Qin,Z., Shi,S., Xu,J. And Fu, H. (2006) ‘Method of License Plate Location Based on Corner Feature’ Proc., Sixth World Congress on
Intelligent Control Automation, Vol. 2, pp. 8645-8649.
[20] Rajeev Ranjan and Abhinav Sinha ‘Automatic Vehicle Registration System for Toll Booths’. We propose the use of image processing
techniques for optimal character recognition in a license plate.
Vol. 2 Issue 1 February 2013
205
ISSN: 2319 – 1058