Simulation conditions: The simulation conditions
considered during the simulation of the presented scheme
are as follows.
i). Input image must have a recognized format i.e.
(.jpg, .tif, .png, .gif etc).
ii). Input image must not be captured totally in front of
camera i.e. not at a great angle to the camera.
iii). The number plate must not be captured in rotated
direction.
In this paper work a scheme is presented for the
automatic number plate recognition. All the simulation has
been performed on MATLAB R2013a using generalized
MATLAB toolbox and image processing toolbox. The
elements such as noise or environmental changes play an
intruding role in license plate verification. These factors can
make the extracted region of the image incomplete. The
recovery stage is applied to reconstruct the license plate
image before the recognition step is taken. A typical
algorithm used for the recovery process is vertical intensity
projection, by which the average width and height of
alphanumeric characters in the segment are measured, and
according to this information, the whole plate is recovered.
A data set of 7 number plate images has been taken for
evaluation of performance of proposed algorithm as shown
in table1. Although, there are many steps from inputting of
image to number extraction. These steps are shown by
figures below. Fig.4. shows the original input test image.
Fig.5. is the snapshot of number plate after morphological
operations, binarization and thresholding. Fig.6. is the
snapshot of number plate with boundary on selected region
for segmentation. Fig.7. shows segmented alphanumeric and
characters present on the number plate. After this process,
segmented numbers and characters will be tested for
recognition purpose. Table1 is showing number plates with
their respective segmented characters and numbers. Also,
final message box containing a string comprised of an
alphanumeric number.
Fig.4. Capturing of input image.