Special Issue Published in Int. Jnl.

Of Advanced Networking & Applications (IJANA)

Image processing based intruder

Detection using Raspberry PI
S.Adhav
Final year Students, Department of Information Technology, Velammal Engineering College, Chennai.
R.Manthira moorthi
Final year Students, Department of Information Technology, Velammal Engineering College, Chennai.
G.Prethija
Assistant professor, Department of Information Technology, Velammal Engineering College, Chennai.
------------------------------------------------------------------ABSTRACT-------------------------------------------------------------
The current paper is mainly about maintaining a secure environment and also free from thefts that are
happening in our home. The present paper discusses about the detection of intruders with the help of the various
devices and software...OpenCV(open source computer vision) is the major software that is being used in our
present work. For detecting faces we are using various algorithms like Haar cascade, linear SVM, deep neural
network etc.The main method that we have proposed in our work is, if any person comes in front of the pi
camera, first it will look for potential matches that we have already stored in our system If the module finds a
match then it continues to record until any intruder comes. If the face is not recognized then the unknown
person’s face will be captured and a snap shot will be sent to the user’s email. The device is developed using
Raspberry Pi b+ with 1.4 GHz quad core processor, raspberry pi camera module and a Wireless dongle to
communicate with users’s email.

Keywords - OpenCV, Rassberry pi, python

---------------------------------------------------------------------------------------------
I. INTRODUCTION obscuring effects[3]. Efficiency of the framework is
regularly broke down by computing the Face location rate
Real time human identification systems are very for every one of the database. The outcome of Object
recognition frameworks uncover that it very well may be
important for security, surveillance and also for biometric
applications. There are numerous approaches to identify, utilized for face recognition even from low quality
track and recognize individuals in open territories, for pictures and shows astounding execution productivity,
example, air terminals, shopping malls and in zones with monitor threats, and avoid/examine criminal activities
limited access, for example, private workplaces, houses addition to face detection motion detection is always an
and so on. Human recognition should be possible from important requirement...Nowadays, the Closed-Circuit
their fingerprints, face, iris, palm prints, palm veins and so Television (CCTV) observation framework is being used
on. Notwithstanding, for quick and helpful individual so as to keep harmony and give security to individuals.
recognition, still the most appropriate biometric parameter There are a few deformities in the video reconnaissance
is facial data. Distinguishing proof of people by utilizing framework, for example, picture is undefined, oddities
facial biometrics is as yet difficult assignment, because of can't be distinguished consequently, a ton of extra rooms
the variable illumination, changing outward appearances are expected to spare the observation data, and costs
as indicated by inclination changes, head introduction and remain generally high[3]. The haar-like highlights is
posture. utilized for face discovery and HOG +SVM calculation is
Throughout the years, different face recognition utilized for face recognition. So as to accomplish a higher
algorithms have been created. Some face recognition exactness and viability V.[4]Desmukh have utilized
techniquesexamine the geometric highlights of facial OpenCv libraries and python PC languageIn the present
pictures, for example, area and separation between nose, work raspberry pi,raspberrypi camera and OpenCV are
eyes, and mouth. Be that as it may, these techniques are utilized to distinguish the interlopers and furthermore to
delicate to the adjustments in illumination and outward communicate with the client.
appearance. In view of this drawback, the greater part of
the face recognition systems attempt to separate some II. RELATED WORK
complete featuresfrom the first face images for comparing Closed-circuit tv(CCTV): CCTVs have proven to be
and matching. hugely popular for security functions because of their low
MayuriDahake and N.N.Mandaogade have proposed face cost performance and smooth functioning . Surveillance
detection and recognition system that is capable of could be very useful for regulation enforcement to analyze
processing images very fast while acquiring very high true and save you crook sports, for recognizing and monitoring
positive face detection rates[1]. threats. Also, surveillance systems have continually been
Object location additionally assumes a noteworthy job in playing a vital role in managing the burglary instances.
distinguishing human faces. Object recognition These CCTV structures generally tend to monitor any
frameworks have been tried crosswise over different activity continuously. This consequences in excessive
standard face databases, with and without noise and other

Page 283
Special Issue Published in Int. Jnl. Of Advanced Networking & Applications (IJANA)

power consumption and large memory usage. Moreover, Fig 1:Raspberry pi board
it does not give alert on any suspicious activities detected.
Radio Frequency Identification (RFID): It uses radio
waves to routinely identify individual or gadgets. There
are many techniques for identification, however the most
trendy is to store a unique serial number that identifies a
human or object on a microchip that is attached to an
antenna.. The blended antenna and microchip are called
RFID transponder or RFID tag and it works in
combination with an RFID reader. The reader antenna is
used to transmit radio frequency (RF) power. Depending
on the tag type, the energy is harvested by the tag antenna
and used to power up the internal circuitry of the tag. The
tag will then modulate the electromagnetic waves
generated by the reader so as to transmit its data back to
the reader. The reader receives the modulated waves and Fig. 2:Raspberry pi Block Diagram
converts them into digital data.
IV. CAMERA INTERFACE
III. HARDWARE AND SOFTWARES USED The raspberry pi camera module used in this paper is
The Raspberry Pi board[fig:1] is the principal module of shown in Figure. 3. The camera module plugs to the CSI
the image capturing and and processing device. Its most connector on the Raspberry Pi. It's able to deliver clear
important parts consist of: primary processing chip unit, 5MP resolution image, or 1080p HD video recording at
memory, energy deliver HDMI Out i.e.. VGA display, 30fps. The camera module attaches to Raspberry Pi by a
Ethernet port, and USB ports. The Picam is hooked up to 15 pin Ribbon Cable, to the dedicated 15 pin MIPI Camera
Raspberry pi through digital camera port that is present Serial Interface (CSI), which was designed especially for
inside the raspberry pi board.The version that we've got interfacing to cameras. The CSI bus is capable of
used is raspberry pi B+ 2017 which is shown below in extremely high datatransfer rates, and it exclusively
fig.1 .We are also using a wireless dongle which is carries pixel data to the BCM2835 processor.
connected to one of the usb ports available. The
Raspberry Pi is a fundamental embedded machine having
a credit score card-sized board computer systems
developed within the UK with the aid of the Raspberry Pi
Foundation[2]. The Raspberry Pi is primarily based on the
Broadcom BCM2835 gadget on a chip (SOC) which
incorporates an ARM1176JZF-S Core (ARM V6K)seven-
hundred MHz CPU processor, Broadcom Video Core IV
GPU having 17 pins, three.5W of electricity, and 1GB of
RAM reminiscence. This major processing chip connects a
camera and display. The Raspberry Pi layout does not Fig 3:Raspberry pi camera module
include a constructed in hard disk or solid drive, as a
substitute it uses an SD card for booting and long term V. SOFTWARES USED
storage. This board is meant to run Linux primarily based
running systems. Python is the main language utilized in 5.1 OpenCV
raspberry pi .This Raspberry Pi module has a Samsung OpenCV 'open source PC vision library' is an open source
elegance 4 micro SD card preloaded with the Raspberry Pi picture preparing library made by Intel 8109 what's more,
NOOBS (New out of Box Software) bundle, and a printed kept up by Willow carport accessible for C, C++, what's
Micro SD card adaptor. The os that we have used in this more, Python. OpenCV is need a compiler like DevC++,
work is raspian stretch. code squares, visual C++. In this paperutilizes python
dialect furthermore, DevC++ compiler. In OpenCV there
are four modules. Predominantly utilized are CV: primary
OpenCV capacities, picture handling calculations, vision
calculations and highgui: GUI capacities, Image and
Video I/O. Utilizing this OpenCV, we will stack pictures
caught by camera. These pictures are in three
configurations paired picture, dark scale picture and
shaded picture. The hued picture contain R G B with pixel
values containing 0-255. It has profundity of the picture
with 8 bits and 3 channels. For stacking the picture
utilizing OpenCV, the taking after program is utilized[2].

Page 284
Special Issue Published in Int. Jnl. Of Advanced Networking & Applications (IJANA)

5.1 NGROK
Ngrok creates a tunnel from the public internet to a port on
your local machine. You can give this URL to anyone to
allow them to try out a web site you're developing without
doing any deployment. It captures all traffic through the
tunnel. It displays information about the HTTP traffic for
your inspection. Raw request/response bytes, parsed
headers and form data, JSON/XML syntax checking and
more are included. It can also replay requests. By default,
ngrok will use ngrok.com as a third-party relay. This
service is provided at no-cost and without registration but
it is possible to get additional features by signing up in the
service (which is pay-as-you-want kind). However, it is
possible to setup and use its own server. This package Fig 4:Elements of neural networks
installs the client part of ngrok. It can be used directly with A node layer is a row of those neuron-like switches that
ngrok.com service or with your own server if you install turn on or off as the input is fed through the net. Each
the ngrok-server package layer’s output is simultaneously the subsequent layer’s
input, starting from an initial input layer receiving your
data.
VI. METHODOLOGY USED FOR FACE DETECTION

6.1 Deep Neural Network

Neural systems are a lot of calculations, demonstrated
freely after the human mind, that are intended to perceive
designs. They translate tangible information through a sort
of machine discernment, naming or grouping crude
information. The examples they perceive are numerical,
contained in vectors, into which all true information, be it
pictures, sound, content or time arrangement, must be
deciphered.Neural systems help us group and characterize.
You can consider them a grouping and order layer over the
information you store and oversee. They help to assemble
unlabeled information as per similitudes among the model Fig 5: Structure of neural network
sources of info, and they group information when they
have a marked dataset to prepare on. (Neural systems can Blending the model's flexible loads with information
likewise extricate highlights that are nourished to different highlights is the way we dole out hugeness to those
calculations for grouping and order; so you can consider highlights concerning how the neural system characterizes
profound neural systems as parts of bigger AI applications and bunches input.
including calculations for support learning,
characterization and relapse.). 6.3 Haar cascade
Face identification is a PC innovation that decides the
6.2 Neural Network Elements areas and sizes of human faces in subjective (advanced)
Profound learning is the name we use for "stacked neural pictures. It recognizes facial highlights and disregards
systems"; that is, systems made out of a few layers. The whatever else, for example, structures, trees and bodies.
layers are made of hubs. A hub is only a spot where Face recognition can be viewed as a progressively broad
calculation occurs, inexactly designed on a neuron in the instance of face confinement. In face confinement, the
human mind, which fires when it experiences adequate errand is to discover the areas and sizes of a known
boosts. A hub joins contribution from the information with number of appearances
a lot of coefficients, or loads, that either intensify or hose
that input, in this way appointing importance to 6.4 Haar features
contributions as to the errand the calculation is attempting OpenCV's calculation is at present utilizing the
to learn; for example which input is most useful is accompanying Haar-like highlights which contributesto
arranging information without mistake? These info weight the fundamental classifiers:
items are summed and after that the whole is gone through
a hub's supposed actuation work, to decide if and to what
degree that flag should advance further through the system
to influence a definitive result, say, a demonstration of
order. On the off chance that the signs goes through, the
neuron has been "enacted."
Here's an outline of what one hub may resemble

Page 285
Special Issue Published in Int. Jnl. Of Advanced Networking & Applications (IJANA)

7.1 Steps followed

 Firstly, we need to install the raspberry os onto the sd
card
 connect all the components to the hardware
 login to the raspberry using vnc server
 connect the dongle to the raspberry pi

7.2 Command used for setting up the OpenCV and to run

the required python file
Initially,all the OpenCV modules are separately installed
on a virtual environment.so,our first step is to type the
command as given below

Fig 6:Haar Cascade

6.5 Cascade of Classifiers
Rather than applying all the 6000 highlights on a window,
amass the highlights into various phases of classifiers and After getting into the virtual environment,we need to
apply one-by-one. (Ordinarily initial couple of stages will change the directory to where we have saved our python
contain less number of highlights). On the off chance that file.refer the below figure.
a window falls flat the main stage, dispose of it. We don't
think about outstanding highlights on it. On the off chance
that it passes, apply the second phase of highlights and
proceed with the procedure. The window which passes all
stages is a face locale.

Final step is to run the required python code as shown

below.

Fig 7:

VII. RESULTS
The testing is done using raspberry pi platform with the
following specifications:
 Quad core 64-bit processor clocked at 1.4GHz
 1GB LPDDR2 SRAM
 5mp Raspberry pi b+ camera module
 Bootable SanDisk Ultra 16GB micro SD Card

Page 286
Special Issue Published in Int. Jnl. Of Advanced Networking & Applications (IJANA)

FLOWCHART

After the script is loaded the face will be recognized.If the

person’s face is unknown then that persons face will be
sent to the users email.

Fig 8: Flowchart

In our project,the main python file that we will running is

the pi_face_recognition.py.all the execution part happens
in this python file.

7.3 Execution

A photo of the intruder will be sent to the user’s email.

VIII. CONCLUSION
From the current work we can able to develop a cost
effective surveillance system, system which recognizes
face efficiently without any delay and sends the email to
the user immediately,if it detects any intruders.

REFERENCES
[1] MayuriDahake,,N.MandaogadeImplementationofRasp
berryPiforHumanFaceDetection&Recognition.
International journal of science and research,23 march
2017
In this figure the python script is loaded. [2] Harshal V. Khodaskar1, Shashank
Mane2International Journal of Advanced
Engineering, Management and Science
(IJAEMS),jan 2017

Page 287
Special Issue Published in Int. Jnl. Of Advanced Networking & Applications (IJANA)

[3] Huu-Quoc Nguyen, Ton Thi Kim Loan, Bui Dinh

Mao and Eui-Nam Low Cost Real-Time System
Monitoring Using Raspberry Pi,2015
[4] Shrutika V. Deshmukh1, Prof Dr. U. A. Kshirsagar2.
Face DetectionandFace Recognition Using Raspberry
Pi,International Journal of Advanced Research in
Computer and Communication Engineering,april 2017
[5] OlegsNikisins, RihardsFuksis, ArtursKadikis,
ModrisGreitansFace recognition system on
RaspberryPi.2015

Page 288