You are on page 1of 9

ABSTRACT

Home security system is needed for occupants convenience and safety everyone
wants to be as much as secure as possible. The approach to home security system
design is almost standardized nowadays. In this paper, we have tried to increase
these standards by combining new design techniques and developed a low cost
home automated security systems. The scope of this project is to design and
develop a circuit that uses a microcontroller embedded with MATLAB that
provides an automatic door access using Artificial Neural Network (ANN). Using
the microcontroller (AT89s52), we are designing a circuit using MATLAB coding
in such a manner that enables a person to open the door using the face. Ripplet
transform is used in taking image and then mean and standard deviation values are
calculated using power law transform. This approach combines frequency
domain approaches to take the system efficient. Power law
transform shows true enhancements of facial features. The
potential of the system is till under improvisation.

INTRODUCTION
1.1. GENERAL
Face detection has always been one of the most interesting and challenging field to
work on. Till now various techniques or approaches have been proposed like
detection as well as recognition from a complex scene by detecting a circular
boundary. Their nose radial grid map was placed as one of the feature to detect the
face In this paper a face recognition system is proposed. This involves ripplet
transform and power law transform. Power law transform is used to enhance the
image whereas Ripplet transform is used to calculate standard deviation and mean
of the region of interest. This dataset is treated as a feature to be trained by
artificial neural network towards face recognition.
In context to image processing representation of images and
signals is of utmost importance. For this purpose several
transforms are used in image pre-processing.
Wavelet transformshows good performance in representing 1D
singularity whereasridgelet and curvelet transform shows better
result on 2Dsingularity representation.
Further the capacity of representation of2D singularity, a new
transform called Ripplet transform (Type 2)has been introduced.
Ripplet 1 transform is the generalization ofcurvelet transform with
two extra parameters c (support) and d(degree). In fact ripplet 2
transform is same as curvelet transformwith values c=1 and d=2.
In

ripplet

singularities.

transform

collection

ofcurves

represents

2D

Moreover Ripplet 2 transform isused for extraction of features as


it is highly efficient inrepresenting textures and edges. Ripplet
transform
restoration,

also

holds

image

latentqualities

compression

in

context

andimage

to

image

de-noising.

The

directionality of a ripplet function is directlyproportional to the


resolution, higher the resolution higher will beits directionality.
Ripplet transform approximates the images oneafter the other
from rough to fine. It also exhibits the property ofanisotropy by
the virtue of which singularities are captured alongvarious curve.
Ripplet transform also has ability to concentrateenergy more than
any other transform.

EXISTING SYSTEM
2.1. EXISTING
The face recognition and detection process is implemented
by

modifying principal component

analysis (PCA)

approach to fast based principal component analysis


(FBPCA) approach, by which the captured image is
detected using a web camera.
Home security system is needed for
occupants' convenience and safety.

2.2. DRAWBACKS:
The reliability and robustness in both the recognition and
detection process is the major drawback.
For an automatic door lock the identification of the right
person is important , so the existing system should be
developed.
Mathematical calculations are not done.

METHODOLOGY
4.1. WORKING
Image of the person is captured and mean, standard deviation is been
calculated.
The calculated values are been compared with the already stored images.
Matched values leads to the opening of automatic door lock.
If values are not matched then the alert is been given in the kind of a door
bell.

4.1. IMPLEMENTATION
Implementation includes:

Face Image Acquisition


Conversion to resize and Grayscaale
Application of Power Law Transformation:
Region of Interest (ROI) Selection:
Face Detection using Feed Forward Neural Network

4.1.1. FACE IMAGE ACQUISITION


Digital test images have been furnished as an input
image. Fig1 Show a single test image to elaborate the
process.
4.1.2. CONVERSION TO RESIZE AND GRAYSCALE
After getting the input then it will be converted into grayscale.

4.1.3. APPLICATION OF POWER LAW


TRANSFORMATION:
Then Power Law Transformation is applied on the image for
enhancement. Then the output becomes.

4.1.4. REGION OF INTEREST (ROI) SELECTION:


Then crop the region of interest from the enhance image.
4.1.5. FACE DETECTION USING FEED FORWARD

NEURAL NETWORK
Generally artificial neural networks are computer implementation
of human brain neuron system. It is a multilayer approach, which
contains one input layer one hidden layer and one output layer.
All the layers may contain several neurons which are nothing but
source of information compilation in a specific manner.

Face Image Acquisition


Conversion to resize and
Grayscale
Application of Power Law
Transformation
Region of Interest (ROI)
Selection
Face Detection using Feed
Forward Neural Network

Flow chart 2 : Implementation Process

4.3. STAGES
To implement neural network of back propagation feed forward
type three stage architecture is needed. This is shown below
Training of Neural Network
Testing of Neural Network
Performance Evolution of Neural Network.

You might also like