Vertical Edge Detection for Car License Plate Recognition

Abstract

There are various methods for recognizing the license plate of cars. In this project, a novel method for car-license plate recognition is used and presents different contributions. The number plate is initially detected precisely on the basis of edge detection algorithms. Next the detected number plate image is fed as input to the recognition module. The first step is that the input detected image is binarized using thresholding technique. It is one among the simplest segmentation processes. In the second step feature extraction is performed on the thresholded output. The objective of feature extraction is to capture the essential characteristics of the characters obtained. Feature extraction is performed based on the parameters such as euler number, major axis length, minor axis length, row and column features, convex area, orientation, filled area etc. Next the feature extracted output is trained using appropriate neural networks .Finally the trained outputs are classified and the exact number is displayed in the user interface platform. The advantage of our method is that even very low resolution images taken by a camera can be processed. Different license plate images under different conditions are given as input and the results are checked. The results show more or less accurate performance and faster processing time.

Foreground Segmentation for Live Videos by Texture Features

“…Figure A2 shows the output produced by our system for the same input frame used by [11]. Similar to the CSVMs misclassification occurs in the initial frame.…”

“…Here each incoming frames are compared with a fixed data set and the images are segmented into foreground and back ground images. [11] used a vertical edge detection scheme to extract the numbers alone from the license plates. Here numbers were treated as the foreground and the other regions were treated as background.…”

A Framework in Shadow Detection and Compensation of Images

“…[16,17] The input of the thersholding function is a gray scale image which is derived from the above step. The output is a binary image representing segmentation.…”

“…These methods have certain restrictions and the results are not produced efficiently. [15,16,17] The shadow detection and compensation process were done by using Dual-Pass Otsu Method. The pixel value is separated in to high and low level intensities.…”