Supervisor: PhD, Associate Professor Leyla Muradkhanli
Keywords:
edge detection, segmentation, kernel, dilation, erosion
License Plate Recognition (LPR) system as known as Automatic
Number Plate Recognition (ANPR) is the action of taking pictures of the
license plates and transforming this optical data into the digital one by using
image processing techniques in real-life time. The LPR systems consist of
THE 3
rd
INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS
dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
169
three standard phases which are number plate detection, plate character
segmentation and character recognition. The detection of a number plate is
a difficult operation due to wide variety of number plate formats and the fact
that ambient variables cause problems during the image acquisition stage.
Accurate character segmentation and recognition mainly rely on the effective
plate detection. In order to locate and isolate the license plate and the
characters, image processing techniques such as edge detection,
thresholding, and resampling are used. LPR systems can be used to tackle
several problems including vehicle identification, entrance admission to
parking lots, security and airport cargo control, road traffic control, and radar
systems.
The LPR system consists of two main parts which are hardware and
software. The hardware part is the cameras which should combine Optical
Character Recognition (OCR) and Neural Network in order to make the
cameras to be able read the license plate of an object or a vehicle’s number
plate and transform the data into the digital one in short time. The software
behind the system can be written in different programming languages such
as Python, C++ (with OpenCV library), MATLAB. The software consists of
four main stages which are localization, segmentation, identification and
regionalization. These stages are processed through the three steps: (I) Pre-
processing, (II) Localization, (III) Character segmentation and recognition.
The first step includes three stages including converting images to
grayscale, binarization and morphological image processing consisting of
non-linear operations related to the shape or morphology of features of an
image with the help of the structuring element as known as kernel. The main
morphological process behind the software is dilation and erosion. The
erosion process causes the image to shrink or decrease in the size; however,
the dilation process makes the image dilate or grow in size. The shrink or
increase size depends on the structuring element. Here are some examples
of dilation and erosion (figure 1).
Figure 1.
Dilation and erosion examples
The localization of the license plate on an eroded image is one of the
most crucial steps of the LPR system. This process can be done by using
the image segmentation which is typically used in order to detect the objects.
However, there can be some problems such as when the background color
|