number plate recognition has been an ongoing innovative research for over the
last few years. Many researches have been carried out to identify the different
type of vehicle such as a truck, car, bus or any four wheeler vehicle. In paper
14, the soble filter algorithm was used to address this issue to get the
edges of the vehicle which is applied to detect and recognize the type of vehicle. The model of 11 the vehicle find out
through in the use of SVM (Support Vector Machine) and contourlet Transform.
They showed many numerical results on data set of pictures. However, they could
not be applied the any technique to real-time capture of video stream 2. In
the paper 16 monocular images technique are used for car recognition. They
applied canny edge detection for detect edges to detect the presence of vehicle
and their number plate and SVM classifier to recognize the vehicle number classification.
In paper 12, recognition of the type of vehicle irrespective of scale, size
and rotation variation of vehicles number plate where 7 we applied the
filter, MACH filter and Log r-theta Mapping techniques.
18, OCR techniques was used 5, which is a commonly used technology for
optical character recognition, which is used for translation of scanned images
of printed text into format of machine encoded text. Basically an OCR technique
is based on neural network fee-forward system. This is proposed for where two
real character images, which is no-overlapping to each other, sets of data uses
for training and training using neural network technology. ANN based neural
network system used for pattern recognition. ANN generally used feed-forward
neural network based intelligent computing architecture, which can be classify
the inputs into a set of target categories. Neural network done work well and
can achieve better performance to other even the size and color of number plate
be different it is also work under in the difficult environment.
The proposed system contains various stages as image acquisition,
pre-processing, number plate localization, character segmentation, character
recognition. The system is designed in Matlab based GUI application.
A. Car Image Captured By Camera
Arduino Uno attached with the motion sensor which detect the motion of car.
Then the motor helps to rotate the camera for capturing the car images from
real time videos 2. We use the high resolution picture quality camera for
image acquisition, identified is captured using high resolution digital camera.
Firstly, we convert the input RGB color image to a gray-scale images. To
speed up the process, the image is first downscaled to 50% of the original.
Here mathematical morphology 6 is used to detect the plate region and the
sobel operator are used for calculation of the threshold value. After this
system we will get a dilated image. Then we use imfill function for fill the
holes, so that we can get a clear binary image.
Figure 2: A GUI layout System overview of Indian Number Plate Recognition
C. Indian Number plate Localization
Pre-processing is the important technique to filtering and edges detection.
The image is pre-processed, passing through gray scale filter and edge
detection method is applied. Which applied to the isolate of the plate region
of interest. Localization 1, 4 is an algorithmic function for identifying a
number plate. By the use of localizing determines the aspect ratio of number
plate of vehicle image. This algorithm search the similar background colour of
in image unified proportion and mean contrast differentiate number plate
objects on a vehicle.
(i) Edge Detection
There are many methods of which performing edge detection of image. We
detected the edges of input image 6, here we using canny edge detector which
used to takes a gray-scale image as its input of this system, and then returns
a binary image of same size 11 as an output image, where the edge detection
function finds the edges in the input gray image.
(ii) Character Segmentation
In the identified number plate region where character are segmented using
function of region-props of Matlab, It is use to find the boxes bounding for
each characters. This function returns the smallest bounding box, which 13
contains a character. So, we can use this method for obtain the bounding boxes
of all character in vehicle number plate.
D. Character Recognition
Template matching is a technique of character recognition. It is method of
finding the fixed location of a template (sub-image), which inside of captured
image. Template matching having
similarities 12 between a given template image and windows with same size in
an image and that identifying the window, that produces highest similarity
measure. It works with pixel-by-pixel comparison and each possible pixel
displacement of the template image.
E. Identify the Stolen Car
The main purpose of this state that recognize and classification of binary
images that have contains character which is received by previous stage. After
doing this stage each character must have a valid label and having an error
factor. If this error factor is greater than a predefined data value will be
used for reject the false characters which passed from previous one. For the
time of classification step some features must have collected data from the
characters. This is use for because image to text into characters conversion.
In number plate each individual character match from the completer alphanumeric
database using template matching method. The matching process checked the template image to all
possible positions in a input larger image and computes a particular numerical
index which indicate that how the template is matches and what quality of
matches the image in that position. After template is matched we having a valid
vehicle number which is matches from the database of excel sheet automatically.
If the stolen vehicle is detect then the alert system will be on. We can be
recognize data using the internet. We store stolen vehicle database in online
database and connect to Matlab java connecter.