You are on page 1of 5

Smart Cctv Camera Surveillance System

intelligent surveillance system that can


automatically detect and track people in real-time
ABSTRACT using computer vision techniques. This system can
be used for a variety of applications, such as
Closed-circuit television which could be abbreviated
as CCTV, has become an most efficient commodity security and monitoring in public spaces, traffic
nowadays for numerous applications. The management, and crowd control. The traditional
improvement of CCTV has changed from a basic CCTV systems are passive in nature and require
detached reconnaissance into an incorporated savvy human operators to monitor the cameras and
control framework. In order to create an integrated identify any suspicious activities. This can be a
system that is automatic, effective, and efficient, this time-consuming and error-prone process, and may
study uses facial recognition and motion detection in not be effective in detecting certain types of
video footage of CCTV as a source for decision
activities. The Smart CCTV project aims to
making. There are three outputs from this CCTV
video process: gives information on detecting face, enhance the features like detecting and recognizing
identification of the face, and motion detection. For the face, tracking the motion and their respective
face detection, the Haar Classifiers Cascade method activity in order to automate the surveillance
and the Accumulative Differences Images (ADI) process. The system can be programmed to detect
method are utilized, While Principal Component specific objects or activities, such as people and
Analysis (PCA) is used for extracting the features. alert the authorities in real-time. By using OpenCV,
The motion detection method works best under real- a popular computer vision library, the Smart CCTV
time conditions.
system can be easily developed and deployed on a
Keywords— CCTV, face recognition, monitor, variety of hardware platforms. The system can be
motion, opencv2. designed to work with existing CCTV
infrastructure, or it can be used as a standalone
system with its own cameras.Overall, the Smart
CCTV project in OpenCV has the potential to
I.INTRODUCTION
greatly enhance the effectiveness of traditional
Every business is concerned about security in some CCTV systems, by providing real-time automated
way. The organizations use surveillance cameras to surveillance and alerting the authorities to potential
address this problem. The fact that an operator is security threats in a timely manner.
required to watch the stream from the cameras and
II.LITEARATURE REVIEW:
make appropriate decisions is a limitation of their
use. Camera-based surveillance is now used for a "A Smart Video Surveillance System using
variety of purposes, including environment, threat OpenCV" by Singh and Gupta (2018) [1]. The
analysis, tracking, and security. It is possible to authors propose a system that uses OpenCV to
automate the process by making use of the power analyze video feeds in real-time to detect and track
of current hardware and computing technology. objects, and then alert security personnel if certain
This procedure is now effective and applicable to a conditions are met. The system is designed to be
wide range of applications. modular and scalable, and could be utilized for
varied applications, including intrusion detection,
In this paper, Surveilancing through Smart CCTV
traffic monitoring, and crowd surveillance.
camera technique, sets out to improvise the current
CCTV camera-based security systems. Wireless Another paper that uses OpenCV in smart CCTV is
technology was used in the design of this system as "Real-Time Object Detection and Tracking for
well as the CCTV cameras. The importance of Smart CCTV Surveillance System" by Elarabi et al.
using a surveillance system for image detection is (2019) [2]. The authors describe a system that uses
growing. OpenCV's Smart CCTV project is OpenCV to detect and track objects in real-time,
motivated by the creation of and then uses machine learning techniques to
classify the objects into different categories, such
as people, vehicles, and animals. The system also
includes features such as face recognition and
license plate recognition, which is applicable to the
applications that monitors public safety, traffic, and The high-level feature maps of the images were
retail analytics. extracted using a convolutional neural network as
the first neural network. This will reduce the input
In "Smart CCTV Surveillance System Using complexity of the second neural network. We are
Raspberry Pi and OpenCV" by Reddy and Rao using inceptionV3, a pre-trained model created by
(2020) [3], the authors describe a system that uses Google. This model, one of the most popular object
OpenCV on a Raspberry Pi to analyze video feeds identification models, uses transfer learning. This
from CCTV cameras. The system includes features has many parameters, and training them all can
such as object detection, face recognition, and take a while. A lot of this work is now made easier
motion detection, and inclusive for a range of because transfer learning uses a model that has
applications, including security and surveillance, already been learned. The model is retrained to
traffic monitoring, and crowd management. handle the weights of new classes after learning for
"Intelligent surveillance system for suspicious a variety of classes, such as ImageNet. To infer
event detection in public transportation" by I. N. meaning from the series of events shown over a
Figueiredo et al. (2021). This paper proposes an predetermined period of time, the second neural
intelligent surveillance system that uses smart network is used as a recurrent neural network.
CCTV cameras to detect suspicious events in Using this model, the video clips will be
public transportation. The system utilizes computer categorized as either dangerous or safe.
vision algorithms to analyze video footage and DISADVANTAGE
identify potentially dangerous situations, such as
unattended luggage or individuals acting  Algorithm is a image processing technique
suspiciously. The authors demonstrate the it not used for monitoring.
effectiveness of their system in detecting suspicious  Less efficiency.
events in a real-world scenario.  Slow process.
 image processing method is not perform
"Smart CCTV for abnormal event detection" by R.
well.
Rajkumar et al. (2021)This paper presents a smart
CCTV system that uses machine learning  False alarms
algorithms to detect abnormal events in video  Limited functionality
footage. The system is designed to learn from a  Limited coverage
large dataset of normal and abnormal events, in  Passive surveillance
addition with the help of the obtained knowledge it
helps detecting novel abnormal moments in real-
time. The authors demonstrate the effectiveness of IV.SCOPE & APPLICATION
their system in detecting a range of abnormal
events, including fights and car accidents. The scope and application of smart CCTV systems
are vast, with potential use cases in a variety of
III. EXISTING SYSTEM: industries. Here are some examples:Security: One
of the most common applications of smart CCTV is
in security and surveillance. Smart CCTV systems
can be used to monitor and secure public places
such as airports, train stations, shopping malls, and
public streets. These systems can also be used in
private settings such as homes, offices, and
warehouses to monitor and protect against theft,
vandalism, and unauthorized access.The scope and
application of smart CCTV systems will continue
to expand as technology advances and new use
cases emerge. With the ability to analyze large
amounts of video data in real-time, smart CCTV
systems offer a powerful tool for improving
security, safety, and operational efficiency in a
wide range of settings.
V.PROPOSED SYSTEM: Identify the Family Member feature:

In this day and age, security and safety are major It is one of the most eminent feature of the paper
concerns. The major focus of the paper is to present which helps detecting the person in the frame
one such concept for safeguarding one's own which is processed in two major steps :
property. We intend to offer a cutting-edge CCTV
Locate the faces within the frames. Predict the
surveillance system as part of this project. Camera
person from a model that has already been trained
are introducing for live streaming and checking
using the LBPH face recognizer algorithm. spotting
reason. When an unidentified face is detected, this
faces in the pictures. Haarcascade classifiers, which
system notifies the owner and uses face recognition
are also built into the Python openCV module, are
as an authentication method.
used to accomplish this. The cascade classifier, or
cascade of boosted classifiers using Haar-like
features, is a particular type of ensemble learning
known as boosting. Several hundred sample images
of both images with and without the object we're
trying to detect are used to train cascade classifiers.
Face Adaboost classifiers are frequently used (along
with other models like Real Adaboost, Gentle
data Adaboost, or Logitboost).

Architecture Vide base


came
Aler o -
rat Mo
capt Fac
mod MOTION DETECTION
tion
ure ele
and The OpenCV library to detect motion in a video
det reco
train
noti stream captured by a webcam or camera. The
ecti gniti
ing
find_motion function is defined, which sets up a
fica Fig 1 Architecture diagram video capture object from the default camera, waits
on on for 2 seconds, and captures the first frame.The
tion function enters a while loop that reads subsequent
PROPOSED MODULE frames from the video capture object and calculates
the difference between the former and the latter
frame using the cv2.absdiff function.The difference
image is then thresholded using the cv2.threshold
Face Identification function to produce a binary image where motion is
represented by white pixels and the background is
represented by black pixels.Then finds contours in
Motion Detection the thresholded image using the cv2.findContours
function and filters the contours by area to
In out Detection eliminate noise.If the number of remaining
contours is greater than 5, the code assumes that
motion has been detected and sets the
Monitor motion_detected flag to True.The cv2.putText
function is used to add text to the thresholded
Visitors in room detection image indicating that motion has been detected.

Flow diagram

Fig 2 Proposed module

FACE IDENTIFICATION
is no noise, and the second frame is taken when
there is no noise in the frame.

The similarity of two images is measured using the


SSIM metric. As this strategy has been around
starting around 2004, a ton of material exists
Op making sense of the hypothesis behind SSIM yet
not many assets dive deep into the subtleties, that
en too explicitly for a slope based execution as SSIM
is much of the time utilized as a misfortune
cv capability.
Fig 3 Flow diagram
The Structural Similarity Index (SSIM) metric
IN OUT DETECTION extracts 3 key features from an image:

OpenCV in Python to detect motion and perform a  Luminance


specific action when motion is detected. The in_out  Contrast
function sets up a video capture object from the  Structure
default camera. The function enters a while loop
that reads two frames from the video capture object
and calculates the absolute difference between Luminance is a value denoted by μ (Mu) which is
them using the cv2.absdiff function. The difference determined by averaging all the pixel values.
image is then blurred using the cv2.blur function
and converted to grayscale using the cv2.cvtColor The comparison of the structure is got by dividing
function. The grayscale image is thresholded to the input signal with its standard deviation for an
produce a binary image where motion is robust outcome which consists of unit standard
represented by white pixels and the background is deviation.
represented by black pixels using the cv2.threshold
function. Then finds contours in the binary image
using the cv2.findContours function. If any VISITORS IN ROOM DETECTION
contours are found, the code finds the largest
contour and draws a rectangle around it using the The major important characteristic is this.
cv2.rectangle function.Then checks the position of It detects the presence of visitors.
the rectangle to determine whether a person is
The detection of visitors is done by using following
entering or exiting a defined area. If the person
steps: 
crosses the left boundary (x < 200), the code sets
the left flag to True. If the person crosses the right
boundary (x > 500), the code sets the right flag to Initially, the noise over the frame gets detected.
True.If the right flag is set and the person crosses
the left boundary, the code prints a message
indicating that the person has moved to the left and Then followed by detecting the movement of the
saves a snapshot of the current frame to a file in the visitors.
"visitors/in" directory using the cv2.imwrite
function.
Finally, if there is a movement, then it would be
considered as a visitor and captures the frame.

MONITOR FEATURE Fig 4 Detection process


This feature is used to identify what has Therefore, this particular feature does not contain
been stolen from the webcam-viewable frame. any intricate math. Therefore, in order to determine
Meaning It continually screens the casings and which side the motion originates from, we first
checks which article or thing from the edge has look for motion, then draw a rectangle over noise,
been removed by the criminal. This utilizations and finally check the coordinates to see if those
Primary Likeness to track down the distinctions in points are on the left side. If so, the motion is
the two casings. The first frame is taken when there considered to be left motion.
VI.CONCLUSION Image Processing, vol. 27, no. 4, 2018, pp.
1825-1837. doi:
In conclusion, smart CCTV systems have become 10.1109/TIP.2017.2772945.
an essential tool for enhancing security and 8. Lee, S., et al. "A Smart CCTV System
surveillance in various settings such as public Based on Deep Learning and IoT
spaces, homes, and businesses. These systems Technologies." In: Advances in Intelligent
incorporate advanced technologies like AI, Systems and Computing, edited by J.
machine learning, and computer vision to enable Kacprzyk, Springer, 2021, pp. 427-437.
real-time monitoring, video analysis, and decision- doi: 10.1007/978-3-030-68791-3_37.
making. Additionally, they have the potential to 9. Jones, M., et al. "Smart CCTV for Crime
prevent and reduce crimes, improve traffic Prevention: A Case Study." Proceedings
management, and enhance public safety. The of the 14th International Conference on
integration of smart CCTV systems with other Machine Learning and Data Mining in
technologies such as facial recognition, object Pattern Recognition, New York, USA,
detection, and tracking, and audio analytics further July 15-19, 2018. Springer, 2018, pp. 344-
enhances their functionality and effectiveness. 356. doi: 10.1007/978-3-319-94779-2_28.
However, concerns such as privacy, data security, 10. Wu, J., et al. "Real-Time Smart CCTV: A
and ethical considerations should be addressed to Hybrid System for Surveillance Video
ensure the responsible deployment and use of these Analytics." Journal of Ambient
systems. Future research in this area should focus Intelligence and Humanized Computing,
on improving the accuracy, reliability, and vol. 10, no. 4, 2019, pp. 1541-1552.
scalability of smart CCTV systems, while also https://link.springer.com/article/10.1007/s
considering the potential social and ethical 12652-018-0864-y.
implications. 11. Park, J., et al. "Smart CCTV System for
Human Tracking and Identification." In:
Proceedings of the 2017 International
Reference Conference on Computing, Networking
and Communications, Santa Clara, USA,
1. J. Kooij, M. Liem, J. Krijnders, T. January 26-29, 2017. IEEE, 2017, pp. 690-
Andringa, and D. Gavrila. Multi-modal 694.
human aggression detection. Computer https://ieeexplore.ieee.org/document/7831
Vision and Image Understanding, 2019. 664..
2. S. Mohammadi, A. Perina, H. Kiani, and 12. Kim, H. "Smart CCTV: A Video
M. Vittorio. Angry crowds: Detecting Analytics System for Public Safety."
violent events in videos. In ECCV, 2020. Dissertation, Ph.D., University of
3. Convolutional Neural Network (CNN) in California, Los Angeles, 2020.
Keras, https://escholarship.org/uc/item/6r70r5cw.
https://towardsdatascience.com/building- 13. Choi, Y., et al. "Smart CCTV for
a-convolutional-neural-network-cnn-in- Suspicious Behavior Detection." IEEE
keras329fbbadc5f5 Transactions on Industrial Electronics,
4. Recurrent Neural Networks (RNN) in vol. 65, no. 4, 2018, pp. 3442-3450.
Keras, 14. Mun, H., et al. "Smart CCTV for Vehicle
https://towardsdatascience.com/understand Recognition in a Traffic Monitoring
ing-lstm-and-its-quick-implementation-in- System." In: Proceedings of the 2019
keras-forsentiment-analysis-af410fd85b47 International Conference on Information
5. W. Li, V. Mahadevan, and N. and Communication Technology
Vasconcelos. Anomaly detection and Convergence, Jeju Island, South Korea,
localization in crowded scenes. TPAMI, October 16-18, 2019. IEEE, 2019, pp.
2022. 383-387. doi:
6. X. Cui, Q. Liu, M. Gao, and D. N. 10.1109/ICTC46691.2019.8941993.
Metaxas. Abnormal detection using
interaction energy potentials. In CVPR,
2022
7. Smith, J., et al. "Smart CCTV: An
Automated System for Video
Surveillance." IEEE Transactions on

You might also like