Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

T-Secured: Intruder Detection in Home Security

Systems
Anna Roy Ankit Suresh Julia Anna Saju
(Computer Science Engineering) (Computer Science Engineering) (Computer Science Engineering)
MBCET (of Aff.) MBCET (of Aff.) MBCET (of Aff.)
Trivandrum, India Trivandrum, India Trivandrum, India
annaannu252000@gmail.com ankitsuresh001@gmail.com juliaannasaju1010@gmail.com

Linda Maria K J
(Computer Science Engineering)
MBCET (of Aff.)
Trivandrum, India
mlindakj@gmail.com

Abstract—Image processing has made a huge impact in the the other two algorithms, the LBPH method is more versatile
advancement of home security.Face recognition, object identi- since it can distinguish both the front and side faces.
fication, and object tracking, among other things, all help to
improve the situation. The existing security systems do not notify II. P ROPOSED M ETHOD
the owners of any potential incursion. They only capture security
video for future reference and do not participate in active threat A. Work flow of face recognition system
prevention.The work uses Local Binary Patter (LBP) algorithm
and histogram equalization to improve the detection rate and
reduce errors in face detection.For future reference, a database is
kept track of all the persons who enter and depart the residence.

Index Terms—Face Recognition, Features Extraction, LBPH


algorithm, Intruder Detection

I. I NTRODUCTION
Face recognition technology is a significant research project
in the fields of computer vision and pattern recognition. It can
detect people’s identities and other information based on the
visual aspects of a face image, and it has a lot of potential.
It is frequently employed in areas such as authentication,
criminal investigation, video surveillance, robotic intelligence,
and medical science. It has a wide range of applications as
well as economic value. When compared to other biological
traits, face features have the characteristics of being nice,
straightforward, and convenient. As a result, users are more
accepting of facial recognition.
Many researchers have developed a variety of face recog-
nition algorithms over the years, including the Sparse Cod-
ing (SC) algorithm, Local Binary Pattern (LBP) algorithm,
Histograms of Oriented Gradients (HOG) algorithm, Deep
Convolution Network algorithm, Linear Discriminant Analysis
(LDA) algorithm, and Gabor feature algorithm, among others.
Eigenfaces, Fisherfaces, and Local Binary Pattern Histogram
are three built-in face recognition algorithms in OpenCV, an
open source computer vision library (LBPH). In comparison to

Identify applicable funding agency here. If none, delete this. Fig. 1. Face recognition system work flow
The input gathering module, feature extraction module,
classification module, and training classifier database module
are the four essential components of most face recognition
systems. The learning acquisition module’s picture informa-
tion, which will be utilised as a test sample for analysis. The
feature extraction module extracts and examines features that
potentially represent human identifying information. In the
classification module, the database-trained classifier is utilised
to categorise test samples in order to derive individual identity Fig. 3. Circular LBP Operator
information.
1) Face Detection:
For face detection, we utilised OpenCV, which has a Haar
cascade classifier. The AdaBoost method is used by the Haar
cascade classifier to recognise various face characteristics. It
begins by reading the picture to be identified and converting eigenvalues for each pixel in a picture. These eigenvalues can
it to grayscale, after which it loads the Haar cascade classifier then be combined to generate the LBP feature spectrum. The
to determine whether or not it contains a human face. feature vector for classification in the LBPH method is the his-
If this is the case, it examines the characteristics of the face and togram of the LBP characteristic spectrum. It divides a picture
draws a rectangle frame around the detected face. Otherwise, into several sub-regions, extracts LBP features from each pixel
it moves on to the next image to be tested. in each sub-region, and creates a statistical histogram of the
2) Feature Extraction: LBP characteristic spectrum in each sub-region, allowing each
The LBP operator is used to express a pixel’s contrast infor- sub-region to use a statistical histogram to describe the entire
mation in relation to its neighbours. In the 3*3 window, the picture through a number of statistical histogram components.
original LBP operator is specified. The benefit is that the mistake of the picture not being properly
It compares the grey value of the next 8 pixels to the median aligned with a specified range is reduced.
pixel value as the window’s threshold. The value of pixel 3) Data Set:
position is indicated as 1 if the neighbouring pixel value is
greater or equal to the median pixel value, otherwise it is
written as (0). Equation 1 shows how to define the function. B. Face Recognition Algorithm
It’s seen in the diagram below.
The recognition procedure revolves around face recognition.
Face recognition is a type of computer vision technology
that analyses facial characteristic data to determine a person’s
identity. Face recognition is separated into two components
in a general sense: face recognition matching and detection.
The technique of face recognition is based on a person’s facial
attributes. and either a facial picture or a video stream as input.
Determine whether a human face exists, and if so,
provide information on the position, size, and positioning of
Fig. 2. Original LBP Operator each major facial organ. Based on this data, the identification
elements in each face are retrieved and compared to known
faces in order to determine the identity of each face. Face
To create 8-bit binary integers, 8 locations in the 3*3 neigh- recognition is a type of biometric recognition that comprises
bourhood are compared. The LBP values of the centre pixel of 4 parts: face picture collecting, face image preprocessing,
points of the window are acquired by converting it to decimal face image feature extraction, matching and combining hard
integers, and these values are utilised to display the texture recognition, all of which are implemented using hardware
properties of the region. A new circular LBP operator is used cameras, network lines, and compute clusters. The technique
in the current LBPH method. Figure 3 and equation 2 show of computation is as follows:
how to express it. T = min {T1, T2, · · · , Tn}
Face recognition is based on this premise. Facial recognition
technology is a type of biometrics technology that works by
The grey value GP of the pixel C’s P neighbourhoods, with collecting and preprocessing face information from camera
a radius of R. The grey value pixel value C is GC (xc ,yc ). equipment. Face recognition is a technique for detecting
This approach frees the LBP operator from the constraints of a people’s faces Face detection’s major objective is to collect
given radius and neighbourhood, allowing it to handle a wider data in order to establish whether there is a human face image
range of size and texture characteristics. It calculates the LBP in the image, as well as the image’s size and orientation, and
to segment the detected human facial image into the adult
face region. Face recognition is the final connection, which
extracts facial feature information and picture information to
see if it’s in the repository. If it is, it has matching identification
information; otherwise, no recognition results will be obtained.
1) Face Feature Extraction by LBPH Method:
Local Binary Pattern (LBP) is a basic yet effective texture Fig. 4.
operator that identifies pixels in an image by thresholding
each pixel’s neighbourhood and treating the result as a binary
integer. • It can alternatively be expressed as a 3x3 matrix with each
It was initially described in 1994 (LBP) and has subsequently pixel’s intensity (0255).
been discovered to be a useful texture segmentation trait. On • After that, we need to utilise the matrix’s centre value as
some datasets, it has also been discovered that coupling LBP the threshold.
with the histograms of oriented gradients (HOG) descriptor • This value will be used to define the 8 neighboring new
boosts detection performance substantially. We can represent values.
the facial shots with a simple data matrix using the LBP and • We create a new binary output for each neighbour of the
histograms. Because LBP is a visual descriptor, it may also core value (threshold). We used a 1 for numbers that were
be utilised for face recognition tasks, as seen in the steps equal or greater than the threshold, and a 0 for values that
below. were lower.
• The matrix will now only contain binary values (ignoring
1. Parameters: the central value). Each binary value from each point in the
a. Radius- The radius represents the radius about the centre matrix must be concatenated line by line into a new binary
pixel and is used to construct the circular local binary pattern. value (e.g. 10001101). Note that other experts concatenate
Normally, it is set to 1. the binary values in different approaches (for example, in a
b. Neighbors- the number of sample points required to build clockwise manner), but the end result is the same.
a circular local binary pattern Keep in mind that now the • After that, we transform the binary value to a decimal value
higher the computational cost, the further sample points you and assign it to the matrix’s centre value, which is a pixel
include. Normally, it is set to 8. from the original picture.
c. Grid X- in the horizontal direction, the number of cells. • We get a new image at the conclusion of this technique
The bigger the dimensionality of the produced feature vector, (LBP procedure) that better matches the properties of the
so more cells there are in the grid and the finer the grid is. original image.
Typically, it is set to 8. • Note: Circular LBP is an extension of the LBP process that
d. Grid Y- in the vertical direction, the number of cells The uses a different number of radius and neighbours.
larger the dimensionality of the produced feature vector, the
more cells there are in the grid and the finer the grid is.
Typically, it is set to 8.

2. Training the Algorithm:


We first should train the algorithm. To do so, we’ll want
a dataset containing the faces of the persons we’re hoping
to recognise. We must therefore assign each image an ID
(which might be a number or a person’s name) so that the Fig. 5.
algorithm can distinguish an input image and provide you
with an output. The ID for all pictures of the same individual Bilinear interpolation can be used to accomplish this. If a data
must be the same. Let’s look at the LBPH computational point falls between the pixels, the value of the four closest
stages now that the training set has been built. pixels (2x2) is used to estimate the new data point’s value.

3. Applying the LBP operation: 4. Extracting the Histograms: Using the picture created
The LBPH’s initial computational step is to develop an in the previous phase, we can now divide the image into
intermediate output that best describes the original image by numerous grids using the Grid X and Grid Y properties, as
accentuating the facial features. To do so, the method employs seen in the following image:
a sliding window idea based on the radius and neighbours
parameters.
• Assume we have a grayscale picture of a face. We may extract the histogram of each zone as follows using
• We can acquire a 3x3 pixel window of a portion of this the image above:
image. • Because we’re working with a grayscale image, each
R EFERENCES
Please number citations consecutively within brackets [1].
The sentence punctuation follows the bracket [2]. Refer simply
to the reference number, as in [3]—do not use “Ref. [3]”
or “reference [3]” except at the beginning of a sentence:
Fig. 6. “Reference [3] was the first . . .”
Number footnotes separately in superscripts. Place the ac-
tual footnote at the bottom of the column in which it was
histogram (from each grid) will only have 256 points (0255) cited. Do not put footnotes in the abstract or reference list.
to record the instances of each pixel intensity. Use letters for table footnotes.
• Then, to make a new, larger histogram, we must concatenate Unless there are six authors or more give all authors’ names;
each histogram. In the case of 8x8 grids, the final histogram do not use “et al.”. Papers that have not been published,
will contain 8x8x256=16.384 places. The features of the even if they have been submitted for publication, should be
original image are represented by the final histogram. cited as “unpublished” [4]. Papers that have been accepted for
Performing the face recognition: publication should be cited as “in press” [?]. Capitalize only
The algorithm has already been trained at this point. Each the first word in a paper title, except for proper nouns and
image from the training dataset is depicted by a different element symbols.
histogram. As a result, given an input image, we repeat the For papers published in translation journals, please give the
procedures for this new image, ending in a histogram that English citation first, followed by the original foreign-language
describes the image. citation [?].
• So, all we have to do to identify the picture that fits the
R EFERENCES
input image is compare two histograms and return the image
with the nearest histogram. [1] L. Li, X. Mu, S. Li and H. Peng, ”A Review of Face Recognition
Technology,” in IEEE Access, vol. 8, pp. 139110-139120, 2020.
• To compare histograms (measure the distance between two [2] M. A. Abuzneid and A. Mahmood, ”Enhanced Human Face Recognition
histograms), we can use a variety of techniques, such as Using LBPH Descriptor, Multi-KNN, and Back-Propagation Neural
euclidean distance, chi-square, absolute value, and so on. We Network,” in IEEE Access, vol. 6, pp. 20641-20651, 2018
[3] A. Ahmed, J. Guo, F. Ali, F. Deeba and A. Ahmed, ”LBPH based
may utilise the Euclidean distance (which is well-known) in improved face recognition at low resolution,” 2018 International Con-
this case, which is calculated using the formula: ference on Artificial Intelligence and Big Data (ICAIBD), 2018, pp.
144-147
[4] G. Xiang, Z. Qiuyu, W. Hui and C. Yan, ”Face recognition based
on LBPH and regression of Local Binary features,” 2016 International
Conference on Audio, Language and Image Processing (ICALIP), 2016,
pp. 414-417

IEEE conference templates contain guidance text for compos-


ing and formatting conference papers. Please ensure that all
template text is removed from your conference paper prior to
submission to the conference. Failure to remove the template
Fig. 7.
text from your paper may result in your paper not being
published.
• As a result, the ID from the image with the closest histogram
is returned by the algorithm. The computed distance, which
may be used as a ’confidence’ measurement, should also be
delivered by the algorithm.
• Then we can use a threshold and the ’confidence’ to
determine if the algorithm appropriately detected the image.
If the confidence is less than the stated threshold, we may
infer that the algorithm has effectively recognised it.

CONCLUSION
Security is a major concern in any part of the world. The
homes we live in are supposedly the safest place we know but
in this era we live, no place is safe, not even our homes.
Especially when crimes like house intrusions and break-in
robbery and so on are at its highest, home security is a vital
issue to be taken care of.

You might also like