Fake Number Plate Recognition

Yug Patel 19DIT058, Viraj Pathak 19DIT059

B.Tech Students,Dept. Of IT,Devang Patel Institute of Advance Technology

and Research

---------------------------------------------------------------------------------------------------------------------

Abstract

In smart cities, automatic number plate recognition (ANPR) is a tool for crime prevention and
investigation. With a set shooting angle and illumination environment, it's often used in parking
management systems and toll booths on roads. If the car isn't genuine, finding it becomes a time-
consuming and tough task. The recognised number plate is shown on the graphical user interface, and it is
then saved in a database with the time and date for subsequent use. Eliminating concerns like traffic
violations and improving parking lot security will be beneficial. Computer vision technology is crucial in
this research for detecting moving vehicle number plate characters. Based on the aforementioned system,
a character recognition technique from number plates discovers and classifies between real and false
number plates. The ml algorithms included in our device makes number plate characters easily
identifiable in the project that we provide.

INTRODUCTION

Increased automobile traffic on roadways, as a result of technical improvements, creates a substantial

demand for traffic control and monitoring. Machine learning and image processing are now used by
computers to monitor traffic. It saves manpower while doing onerous jobs like highway vehicle counts,
parking violation warnings, database maintenance, blacklist and stolen vehicle notifications, and so on.

Management of vehicles and transportation is a time-consuming and complex task. There would be a lot
of blunders and obstacles if everything is done by hand. As a result, a recognition system for automatic
number plate detection is necessary.

Residential parking spaces are in short supply, which requires a solid management-driven strategy. A
major portion of the world's population lives in cities, necessitating the usage of secure and convenient
parking places on a daily basis. This system can keep of all permitted vehicles registered in the parking
management system, as well as their users' information. Every time a car joins or exits the parking lot,
real-time data reflecting the entry of these vehicles is generated.
In recent years, computer vision technology has made good advances in addressing real-world issues. As a
corollary, we can anticipate a new era of machine vision applications.

To compensate for the limits and flaws of the licence plate identification strategy, this study presents a
vehicle recognition technique based on the overall visual attributes of the cars. This method can be used
in conjunction with the standard licence plate recognition method to solve the problem of a missing or
erroneous licence plate.

PROPOSED SYSTEM

Using an automatic number plate recognition system, the suggested system will automatically recognise
unauthenticated vehicles. The database would keep track of all authenticated vehicles in residential zones.
When a car approaches the parking system, the sensor detects its existence, the camera captures the
image, and the number from the number plate is extracted.

The transformation of data from video cameras into a new dimension is an attempt to give computers
vision and machine learning the ability to learn activities performed by computers. Many image
processing functions are supplied by the Open cv library, which is compatible with all operating systems.
It primarily focuses on real-time image processing for vehicle number plate recognition.

MACHINE LEARNING METHOD

Linear Regression Learning the Model

Learning a linear regression version consists of time the values of the coefficients with inside the
illustration the use of the facts at hand. We'll study 4 strategies for growing a linear regression version on
this section. This isn't always sufficient to put in force them from scratch, however it's going to provide
you with an concept of the computation and trade-offs involved.

Since the version has been very well researched, there are numerous greater strategies. Ordinary
Least Squares need to be referred to given that it's far the maximum usually used technique in general.
Also maintain in thoughts that Gradient Descent is the maximum usually taught approach in device
mastering classes.
SIMPLE LINEAR REGRESSION

When the usage of easy linear regression with a unmarried input, data may be used to estimate the
coefficients. This necessitates calculating statistical houses consisting of means, trendy deviations,
correlations, and covariance from the results. To traverse and degree data, all the statistics need to be
available.

IMAGE PREPROCESSING

Image Saliency Detection

Image saliency detection is a technique based on imitating human visual observation and ROI (Region of
Interest) extraction. It is a comprehensive process of image analysis, feature extraction and exploration
of human visual characteristics. Saliency detection has great significance for various applications based
on image analysis and processing. In this paper, the saliency areas of training set were extracted in order
to remove the partial background of vehicle image and to highlight the vehicle itself. During the feature
extracting, by filtering out a lot of background characteristics, we can reduce computational complexity,
improve the system's speed and get good training results to extract the saliency areas of the vehicle's
images. First, a given image was decomposed into compact, perceptually homogeneous elements that
abstract unnecessary details. Then compute two measures of the contrast that rate the colour
uniqueness and the spatial distribution of these elements. A saliency measure was derived from the
element contrast. Finally, assign the actual saliency values to the input image to get a pixel-accurate
saliency map.

Background Filtering

By extracting the saliency areas, the background and foreground of the image can be simply
distinguished. The brightness of the saliency areas is significantly greater than that of the background.
By setting the pixel value of the saliency areas to one and the background to zero, we can form an image
matrix which is consisted of zeros and ones. Then, apply dilation operation to the matrix. In the
following step, each pixel value of the matrix is multiplied by the corresponding pixel value of the
original input image to obtain the pre-processed image. After the pre-processing, some of the
background of the training images are eliminated to retain a complete body of the vehicle. This can
avoid extracting a lot of background-characteristics during the feature extraction and contribute to
improve recognition and retrieval precision.
VEHICLE RECOGNITION

Training the SVM Classifier

Our System calls for a pair statistics. System Will accepts the CSV layout for the statistics collection.
Tags File wherein Datasets will offer essentially Hours and Score to make prediction.
There are at least 8 alternative shooting directions corresponding to different view angles for each type
of vehicle in order to adapt to any arbitrary angle of picture shot by the smart phone.
First, for each image in the database, extract the saliency areas. Second, identify and explain the
saliency filtered images' features. Third, create a "visual dictionary" by clustering the feature vectors.
Fourth, use the "visual dictionary" to quantify each aspect as a "visual word." Finally, to turn the image
into a "code book," collect statistics on occurrence frequency for each "visual word."
1000 photos can be sorted into 10 categories by marking the correct kind for each vehicle to make
handwritten annotations for each image. Then, collect the “code books” of the same kind of vehicle
type together as a set and train the set by SVM method to learn a classifier, each classifier for each
vehicle type can be obtained by this method.
CONCLUSION
The experimental results show that using the method in this paper can detect the vehicle face and
foreground images of unlicensed vehicles, license plate covered vehicles, license plate defaced and
forged plate vehicles. The recognition precision is 84%. The research of this paper solved problems of
foreground extraction and fake plate vehicle recognition, lays a strong foundation for the future
development of fake plate vehicle recognition and retrieval system.

REFERENCES
● A. Mutholib, T S. Gunawan and M. Kartiwi, “Design and implementation of automatic number
plate recognition on android platform,” In Computer and Communication Engineering (ICCCE)
● A G.Liaqat, “Mobile real-time license plate recognition,” Ph.D. dissertation, Goteborg: Linnaeus
University, 2011.
● A. Visvanathan and T. Chattopadhyay “Car number plate recognition for smart building
management,” In Digital Signal Processing, IEEE
● F. Perazzi, P. Kr01henbühl, Y. Pritch, and A. Hornung, “Saliency filters: Contrast based filtering for
salient region detection,” In Computer Vision and Pattern Recognition (CVPR), IEEE
● J. Matas, et al. "Robust wide-baseline stereo from maximally stable extremal regions. Image Vis
Comput," Image & Vision Computing
● K. Mikolajczyk, C. Schmid, “An affine invariant interest point detector,” Springer Berlin
Heidelberg.
● L. Juan, O. Gwun, “A comparison of sift, pca-sift and surf,” International Journal of Image
Processing, vol. 3, no. 4
● C. G. Harris, and M. Stephens. "A combined corner and edge detector," Proc of Fourth Alvey
Vision Conference
● W. Pei, Z J. An, Y Y. Zhu, X. Jia, X W. Zuo, F S. Wangl, “A rapid vehicle recognition and retrieval
system,” In Systems and Informatics (ICSAI), IEEE