Volume 6, Issue 6, June – 2021 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Smart Surveillance Using Computer Vision

PANTHANGI SAI LOHITH GUNDOJI SHIVASAI LAGGONI VIGNAN GOUD
Electronics and Communication Electronics and Communication Electronics and Communication
Sreenidhi Institute of Science and Sreenidhi Institute of Science and Sreenidhi Institute of Science and
Technology Technology Technology
Hyderabad, India Hyderabad, India Hyderabad, India

CH Lowkya DR. SI Khan

Assistant Professor, ECE Assistant Professor, ECE
Sreenidhi institute of science and technology Sreenidhi institute of science and technology
Hyderabad, India Hyderabad, India

Abstract:- Security and protection are the most important scalable face recognition. Ready-to-use transmission system.
issues in modern times. This document reviews some of Poor performance in the early stages of operations (especially
the latest developments in the field of computer vision and the high false alarm rate) has led to research on more
image processing to solve safety issues. It also describes complex solutions. Artificial intelligence concepts such as
the efforts of the development teams to incorporate some neural networks and expert systems have been integrated into
of these pioneering ideas into a consistent prototype more computer-oriented systems.
system. Computer vision expands the paradigm of image
processing, including the understanding of scene content, II. LITERATURE SURVEY
object tracking and classification. The development of
computer vision technology often takes into account very  Yong-Liang Zhang, Zhi-Qin Zhang, Gang Xiao, Rui-
specific applications, and the goal of a more complete Dong Wang proposed a method to detect intruders by
understanding of computer vision systems still exists identifying human targets in the images captured by the
outside the existing technology, at least for now. In this surveillance system.
paper we are mainly stressing on the features of smart  Jayathi Ghosh Dastidar, Rana Biswas proposed a method
surveillance which are monitoring feature, noise detection, for intrusion detection in critical environments, by
face identification and visitors in room detection. This can changing the focus of the camera to ensure security and
be achieved by integrating camera with python computer tracking the attacker's movement.
vision algorithms.  Hang Chen zeo, Dong Chen,and Xifeng Wang proposed a
method of controlling the algorithm program of a mixed
Keywords:- Computer Vision, Object Tracking, Computer mathematical model, which is mainly based on the
Vision Algorithms, Face Identification. standard model of Gaussian mixture. After updating the
model according to the characteristics of continuous video
I. INTRODUCTION frames, the background model is divided into static
regions. And dynamic areas and update the background of
In the modern world, the security paradigm has changed different strategies. Penetration is judged based on the
from "investigating incidents" to "preventing potentially position of the target's center of mass, where the center of
catastrophic incidents." Affordable digital video surveillance gravity is in a given space. Means breaking in and calling
systems only provide the infrastructure for capturing, storing, the police.
and distributing video, leaving the task of threat detection
specifically to human functions. Human gazing of III. REQUIREMENTS
surveillance video is a completely human in depth task. It is
largely agreed that looking video feeds calls for a better Since this is software based project, it must run on
degree of visible interest than maximum regular errands. In certain hardware and operating systems, the following
particular, vigilance, that is, the power to remain targeted and requirements must be met to run this software. Any version
react to rare events, is very exacting and is susceptible to of Windows / Linux / Mac OS so that it can run on any
errors because of neglect of attention. Face recognition platform. You need to install Python on your system to run
technology has created nice strides within the past 2 decades. successfully.
Verify credentials for secure transactions, observation and
security tasks, and access management. These applications Python package required are:
usually run in a controlled environment and detection  openCV
algorithms, which can take advantage of the limitations of the  Skimage
environment to achieve a high level of detection accuracy. In-  Numpy
depth research is ongoing in the field of intrusion detection.  Tkinter
In our project, we successfully developed a reliable and

IJISRT21JUN948 www.ijisrt.com 941

Volume 6, Issue 6, June – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
In terms of hardware requirements, you don’t need Below are the different features which can performed
much, but you still need to meet the following requirements: by using this project:
 Working PC or Laptop 1. Monitor
 Webcam with drivers installed. 2. Identify the person
 Flash light or led if using this at night. 3. Detect the noise
4. In and Out detection
IV. METHODOLOGY
4.1 Monitoring feature:
The scope of our project is , generally in normal CCTV, This feature is used to find stolen goods in the frame
it only record the visual content in its vision, where as in this visible to the webcam. Value Constantly monitor the frame
smart cctv , we have introduced features like monitoring the and check which object or thing in the frame has been taken
objects, noise detection , face identification and person in and by the thief. It uses structural similarity to find differences in
out capturing features. two frames. Take two pictures first when there is no noise,
and take the second picture when the noise stops. The
Structural Similarity Index (SSIM) metric extracts 3 key
features from an image: Luminance, Contrast and Structure.
The comparison between the two images is performed on the
basis of these 3 features.

Fig.1: SSIM algorithm block diagram

Fortunately, because of the ski-image package in to divide the image into multiple grids, as shown in the
Python, we don't have to copy all the math into Python, following figure: predict what you want to meet, apply the
because ski-image has a prebuilt function that can do all these same steps to the mark The histogram is compared with the
tasks for us just by calling the built-in functions. We captured trained model. This is how the function works.
them before, so we just need to enter them and get a masked
image with score.

4.2 Face identification feature:

This feature is very useful feature of our project, It is
used to find if the person the frame is known or not. It do this
in two steps:
1. Find the faces in the frames
2. Use LBPH(local binary pattern histogram) face recognizer
algorithm to predict the person from already trained model.

The LBPH uses 4 parameters: Radius, neighbors grid x and

grid y .
The first calculation step of LBPH is to create an Fig.2: LBPH model
intermediate image that better describes the original image by
extracting facial features. To this end, the algorithm uses a 4.3 Noise detection feature:
sliding window concept based on radius and neighboring This feature is used to find the noises in the frames.
parameters. Now that we have the image generated in the Talking in simple way all the frames are continously analyzed
previous step, we can use the parameters Grid X and Grid Y and checked for noises. Noise in checked in the consecutive

IJISRT21JUN948 www.ijisrt.com 942

Volume 6, Issue 6, June – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
frames. Simply we do the absolute difference between two
frames and in this way the difference of two images are
analyzed and Contours (boundaries of the motion are
detected) and if there are no boundaries then no motion and if
there is any then there is motion. As you would know all
images are just integer/ float values of pixels which tells the
brightness of pixel and similarly every pixel has that values
of brightness. So we just do simply absolute difference
because negative will make no sense at all.

Fig.5: Intruder In and out capturing feature

Fig.3: noise detection grid plot

4.4 Intruder In and out capturing feature:

This is the feature which can detect if someone has
entered in the room or gone out.

So it works using following steps:

1. It first detect for noises in the frame.
2. Then if any motion happen it find from which side does
that happen either left or right.
3. Last if checks if motion from left ended to right then its
will detect it as entered and capture the frame. Or vise-
versa.

So there is not complex mathematics going on around in

this specific feature. So basically to know from which side
does the motion happened we first detect for motion and later Fig.6: Object monitoring feature
on we draw rectangle over noise and last step is we check the
co-ordinates if those points lie on left side then it is classified VI. FUTURE SCOPE
as left motion.
Due to technological advancements such as low
V. RESULTS capacity but high computing power, the project can be widely
used. Below are some future exercises for this project. There
is a high-performance device. Additional features, such as
lethal weapon detection, incident detection, fire detection and
more...Build a stand-alone application without requirements
such as Python. Build an independent device. Adding DL
support will expand the scope of the project because we can
use DL to add more features.

VII. CONCLUSION

In-depth research is ongoing in the field of intrusion

detection. In this work, we successfully completed the tasks
set for continuing the project. The purpose is to recognize
faces and process these faces, and identify whether the
recognized faces match existing faces. Exist in the database
(the database contains pre-encoded training faces) If the
detected face does not match the face in the database, will be
sent with frame capture hit the intruder in the face. Although
Fig.4: face identification all these processes run in the background, users can broadcast
the entire live broadcast via their PC or through the Android

IJISRT21JUN948 www.ijisrt.com 943

Volume 6, Issue 6, June – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
application we developed to complement the live broadcast [9]. Junying Zeng, Xiaoxiao Zhao, Chuanbo Qin, “Single
on the smartphone. Even the difference between the most sample per person face recognition based on deep
advanced computer vision standards and the analytical skills convolutional neural network”, 3rd IEEE International
of ordinary security personnel will be seen. However, long- Conference on Computer and Communications (ICCC),
term pressure to improve security and reduce the cost of 2019.
providing surveillance services is increasing, which is still the
driving force for research on advanced computer vision
technology.

ACKNOWLEDGEMENT

As we present our project on “SMART

SURVEILLANCE USING COMPUTER VISION”, we take
this opportunity to offer our sincere thanks to all those
without whose guidance this project might have remained a
dream for us. We sincerely thank the Department of
Electronics and Communication, whose ideas have
specifically guided us to implement this project.

REFERENCES

[1]. Yong-Liang Zhang, Zhi-Qin Zhang, Gang Xiao, Rui-

Dong Wang, Xia He, “Perimeter Intrusion Detection
based on Intelligent Video Analysis”, 15th International
Conference on Control, Automation and Systems
(ICCAS), 2015.
[2]. Jayathi Ghosh Dastidar, Rana Biswas, “ Tracking
Human Intrusion through a CCTV”, International
Conference on Computational Intelligence and
Communication Networks (CICN), 2015 .
[3]. Paul L. Rosin, Tim Ellis, “Detecting and Classifying
Intruders in Image Sequences”, in Proceedings of the
British Machine Vision Conference, 1991 .
[4]. Hang Chen, Dongfang Chen, Xiaofeng Wang,
“Intrusion detection of specific area based on video”,
9th International Congress on Image and Signal
Processing, BioMedical Engineering and Informatics
(CISPBMEI), 2016.
[5]. Jin-xiang Wang, “Research and implementation of
intrusion detection algorithm in video surveillance”,
International Conference on Audio, Language and
Image Processing (ICALIP), 2016 .
[6]. Mozammel Chowdhury, Junbin Gao, Rafiqul Islam,
“Human detection and localization in secure access
control by analysing facial features”, IEEE 11th
Conference on Industrial Electronics and Applications
(ICIEA), 2016 .
[7]. Yang Xiao-Qiang , Zhang Xiao-Bing , “ Face
Recognition Research Based on the Fusion of Layered
LBP Feature”, International Conference on Industrial
Informatics – Computing Technology, Intelligent
Technology, Industrial Information Integration
(ICIICII), 2017.
[8]. Nikita Bakshi, Vibha Prabhu, “ Face recognition system
for access control using principal component analysis”,
International Conference on Intelligent Communication
and Computational Techniques (ICCT), 2019 .

IJISRT21JUN948 www.ijisrt.com 944