Automatic Vehicle Number Plate Recognition

Project : Automatic Vehicle Number Plate Recognition System
Semester : Spring 2011
Platform : Java


Automatic Vehicle Number Plate Recognition System was a semester long team project (team of 3). The aim of this project was to develop a smart and simple algorithm for a generic vehicle license plate recognition system from video. 


Modules :

• Obtaining frames from input video

• Consecutive frames are compared

• If sufficient difference is noted, it is assumed that a moving vehicle is present in that image, and the frame is selected for further processing.

• Extracting number plate from obtained frame

• Convert color image to gray scale image

• Apply Canny edge detection to get edges

• Apply Hough Transformation to get straight lines

• Enhance output and apply Dilation 

• Detect rectangles

• Segmentation of extracted number plate

• Deskew number plate area

• Get segments by finding horizontal boundaries between characters

• Enhance segments by eliminating dots, stretches and redundant space on the sides of characters

• Recognizing characters from segmented images

• Normalize segments

• Template matching using neural network principles

My major work involved extracting number plate from the selected frame.
Here is an example :

Sample Image

Convert to gray scale

Edges using Canny Edge Detector

Straight Lines after applying Hough Transformation

Extracted Number Plate