Second International Conference on Advances in Engineering and Technology
(ICAET-2022)- (Online)
ICAET-2022
Organised by
RSP Conference Hub, Coimbatore, Tamilnadu, India
Date: 29/09/2022 & 30/09/2022
Paper ID :220946 Category : UG / PG / RS / Faculty
Paper Title : INDENTIFYING WOMEN HARASSMENT IN PUBLIC CCTV CAMERAS
USING ARTIFICIAL INTELLIGENCE TECHNIQUES
Presented by
Mr Ajay Vishal R P,
Student – UG
Bannari Amman Institute of Technology,
Erode, Tamil Nadu
Authors Details
Mr Sakthi Aravind J,
Mr Kuralamudhan J
Ms Harsha M R
Ms Pooja Shree C
Student - UG
Bannari Amman Institute of
Technology,
Erode, Tamil Nadu
Paper Id : 220946
Presenter Category : UG
ABSTRACT
3
The general public most likely living in
the most terrible time our cutting edge
society has at any point found concerning
ladies security..Ladies feel hazardous to travel
solo at odd hours. There are numerous
android applications that are
created for ladies' wellbeing. The proposed
framework is an endeavor made to take care
of the issues of ladies wellbeing. Clearly there
will be CCTV cameras in each open spot. In
the event that any uncommon movement of
an individual is perceived by a camera, the
alarm message will be shipped off the close
by police headquarters which makes them
alert. This undertaking is made for the
accommodations, everything being equal.
INTRODUCTION
4
• Among the most terrible nations
in
wrongdoing, India has a despicable
history in
all types of sexual double-dealing.
• The proposed framework is a completely
coordinated progressed AI based framework
which can obviously recognize what is going
on of the casualty by utilizing public CCTV
• Alarm message quickly to the close by
control room or police headquarters to
protect the concerned specialists
There are sure prior CCTV cameras that
really do make an impression on the saved
contacts yet not a solitary one of them is
compelling and adequately fast and as
indicated by an overview, the current
innovation doesn't encourage most ladies.
PROPOSED WORK
5
• Human Detection
• Gender Prediction
• Anomalous Activity
Detection
Despite the fact that there are a number of
particular applications and numerous IOT
based thoughts for ladies security, this
proposed framework requires no
extraordinary equipment to cycle.The
undertaking is ordered into three modules.
WORK FLOW
6
• Step1: Start
• Step 2: Access Live CCTV camera stream.
• Step 3: Detects the Presence of humans
• Step 4: If any human activity is detected it will enter the next module (i.e.) Gender prediction.
• Step 5: It will identify if any women are there or not.
• Step 6: Then it will be looking for anomalous activity.
• Step 7: If any activity is found it will be sending an alert message to the nearby police station or control
room
HUMAN DETECTION
7
• Surface based techniques like Histograms of Oriented
Gradient (HOG) utilize high layered highlights in view
of edges and use Support Vector Machine (SVM) to
identify human districts
• The standardized outcome gives better
execution on variety in light and force.
• The HOG is contrast-standardized, this is finished
by working out force over a bigger region a few
cells known as square, then this worth is utilized
to standardize all cells inside that square.
The human arrangement strategies could be
partitioned into three classes: shape-based,
movement based and surface based.
GENDER PREDICTION
8
• Observation camera the alarm
message will be shipped off close
by police headquarters.
• Making age and orientation assessment from a solitary
face picture a significant assignment in astute applications,
for example, access control, human-PC collaboration,
policing, knowledge and visual reconnaissance, and so on
• Age is one of the vital
facial credits, playing a very
essential job in friendly connections
assuming a female is remaining solitary in a rail
line station or in some other public places
particularly at evening time's is identified under
a live observation
ANOMALOUS
ACTIVITY
DETECTION
9
• The calculation will be searching for an uncommon
movement, for example, running and so forth. Assuming
that any such sort of action is distinguished the message
will be shipped off the close by police headquarters.
• We propose a MIL positioning misfortune with meagerly and perfection imperatives for a
profound learning organization to advance abnormality scores for video fragments.
• We treat all irregular recordings as
one class and typical recordings as
another class.
In the event that expect that, there are numerous
ladies remaining in a typical spot. It will be
considered as an ordinary action and on that
time assuming the message is conveyed to the
close by police headquarters it very well may be
set apart as spam.
ACCURACY
10
0
1
2
3
4
5
6
Category 1 Category 2 Category 3 Category 4
Chart Title
Series 1 Series 2 Series 3
The lower the percentile edge,
the higher the quantity of cells we
group as hazardous and the higher
is our model precision, yet is it great.
We need to track down the ideal
harmony between risk grouping and
model precision to get the best end-
client experience. There is definitely
not a solitary issue that can be more
terrible for an
individual's encounter than trust on
an application that is letting them
know that they are venturing into a
protected zone and end up being a
survivor of wrongdoing.
CONCLUSION
11
A ton of NGOs restoration focuses and helpline
numbers have been made functional in the
previous years however they are largely fixes to
the badgering that has proactively occurred
and not the counteraction that we want. There
are sure previous applications that truly do
make an impression on the saved contacts yet
not a solitary one of them is compelling
and adequately speedy and as indicated by a
study the current innovation doesn't encourage
most ladies.