license plate extraction and character segmentation   by hina kochhar niti goel supervisor...

34
LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava

Upload: dwayne-conley

Post on 18-Jan-2018

264 views

Category:

Documents


0 download

DESCRIPTION

Introduction License plate recognition is an image-processing technology to identify vehicles by their license plates. All vehicles have the identity displayed (license plate) so no additional transmitter or responder is required. These systems are country specific.

TRANSCRIPT

Page 1:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

 

 

LICENSE PLATE EXTRACTION AND

CHARACTER SEGMENTATION 

ByHINA KOCHHAR

NITI GOEL

SupervisorDr. Rajeev Srivastava

   

 

Page 2:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Introduction

• License plate recognition is an image-processing technology to identify vehicles by their license plates.

• All vehicles have the identity displayed (license plate) so no additional transmitter or responder is required.

• These systems are country specific.

Page 3:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

System Definition

• The recognition of a license plate can be divided in three major parts :

Extraction of the license plate Isolation of the individual characters Character recognition

Page 4:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Project Focus

• This project is concerned with the task of extracting the information needed to identify the characters.

• The input is an image of a vehicle assumed to be speeding.

• The goal is to extract the license plate and isolate the characters.

Page 5:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Indian License Plates

• Licence plate number is issued by the district-level Regional Transport Office (RTO) of respective states.

• All license plates are supposed to follow certain guidelines.

• We assume license plates are in the correct format for our project.

SS DD AA NNNN

Page 6:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Plate Extraction

• Assumptions made concerning the shape and appearance of the license plateThe license plate is a rectangular region of an easily

discernable colorThe width-height relationship of the license plate is

known in advanceThe orientation of the license plate is approximately

aligned with the axes

Page 7:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Steps for Plate ExtractionDetection of white pixels using threshold method

Using a sample white image

Candidate area detection

Extracting license plate region from candidate area

Crop license plate

Page 8:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Original Image

Page 9:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Detection of white pixels using threshold value

Finding threshold using color sample• The license plates have a standard white color. • A 5 X 5 sample of this white color is taken. • The average of the matrix corresponding to this sample is

found.• This is used to detect white pixels from the image.

Page 10:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Detecting white pixels in the image • For each pixel in the image, RGB values are compared

with the threshold.• If they lie in the range specified by the threshold, then

that pixel is assigned a value of 0.• The image obtained is subtracted from the original image

to get an image in which the license plate (in RGB format) is detected.

• This RGB image is then converted into grayscale image.

Page 11:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Image after white pixel detection

Page 12:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Image Containing License plate

Page 13:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Candidate Area Detection

Dilation of Image• The basic effect of the operator on the image is to

gradually enlarge the boundaries of regions of foreground pixels.

• Thus areas of foreground pixels grow in size while holes within those regions become smaller.

Page 14:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Finding connected components• Pixel connectivity is used to find all the connected

components in the image.• Finding all connected components in an image and

marking each of them with a distinctive label is called connected component labeling.

Page 15:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Finding parameters of connected components• The connected component labeled image is used to get the area and

location of the components.• Minimum license plate area and height-to-width ratio have been

found. Minimum area = 2000 Minimum Height-to-Width ratio = 0.17 Maximum Height-to-Width ratio = 0.50• The connected components satisfying these values are found.• The connected component having the license plate will be at

maximum depth in the image.

Page 16:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Candidate Region

Page 17:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Extracting license plate region from candidate area

• The candidate area image is now used to find the license plate location by using the same steps as in candidate area detection.

Page 18:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

License plate extraction using same method used in candidate region selection

Page 19:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Crop License Plate

• The sum of the lines and of the columns of the image is computed, obtaining one vector for each direction.

• For these two directions, first point respectively at the left and the right side of the vector that is superior or equal to the average is found, thus obtaining a rectangle to be used for cropping the plate.

Page 20:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Figure showing sum of columns in improved image

Page 21:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Figure showing sum of rows in improved image

Page 22:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Extracted License Plate

Page 23:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Conversion from gray-scale image to binary image and resizing

• Gray-scale image of license plate is converted into binary image

• Background of image is shown by binary value 0 and foreground are shown by binary value 1

• Binary image is resized to 50 X 250

Page 24:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Binary Image of License plate

Page 25:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Resized binary image of license plate (50 X 250)

Page 26:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Character Segmentation

• The extracted plate is divided into nine images, each containing one isolated character.

• Since no color information is relevant, the image is converted to binary colors before any further processing.

Page 27:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Peak-to-Valley Method

• A horizontal projection of a binary image of the plate reveals the exact character positions.

• Changes from valleys to peaks are searched by counting the number of white pixels per column in the projection.

• A change from a valley to a peak indicates the beginning of a character, and vice versa.

Page 28:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Sum of the columns graph of binary resized image of license plate

Page 29:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Images of segmented characters of license plate

Page 30:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Conclusion

The results obtained are as follows :

TASK SUCCESS RATE

Plate Extraction 78.4%Character Isolation 88.9%

Page 31:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Applications

• LPR applications have a wide range of applications which use the extracted plate number :ParkingAccess controlTollingStolen carsPublic parking etc.

Page 32:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

Future Work

• This project isolates the characters of the license plate.

• Further work in this project will include identifying these characters.

• Also the project can be generalized to identify all kinds of plates.

Page 33:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

References[1] A Licence Plate recognition tutorial,

www.licenseplaterecognition.com[2] Indian License plates,

http://en.wikipedia.org/wiki/Indian_licence_plates[3] Sandeep Phukan, "The scent of a scam”, The Week, January 12,

2003.[4] M. Shridhar, J.W.V. Miller, G. Houle, L. Bijnagte, “Recognition of

License Plate Images: Issues and Perspectives”, Proc. of the Fifth Intl. Conf. on Document Analysis and Recognition, pp. 17-20, Sept. 1999

[5] Pixel connectivity, http://www.cee.hw.ac.uk/hipr/html/connect.html[6] Connected component labeling,

http://www.cee.hw.ac.uk/hipr/html/label.html

Page 34:     LICENSE PLATE EXTRACTION AND CHARACTER SEGMENTATION   By HINA KOCHHAR NITI GOEL Supervisor Dr. Rajeev Srivastava        

[7]Remus Brad, “License plate recognition system”, Computer Science Department, Lucian Blaga University, Sibiu, Romania

[8]Cohen, H., Bergman, G., Erez, J., 2002. Car License Plate Recognition, Project Report, Vision and Image Sequence Laboratory, Technion, Israel.

[9] Matlab 7.0 Help