Massive Growth of AI in
Video Surveillance
This Artical published in Safe secure magazine November 2018 edition.
In
June 29, 2001 release A.I. Artificial Intelligence, co-written by American science
fiction drama film, co-produced, and directed by Steven
Spielberg and based on the 1969 short story "Supertoys
Last All Summer Long" by Brian
Aldiss. In the late 22nd century, rising
sea levels from global warming have wiped out coastal cities such as Amsterdam, Venice,
and New York, and drastically reduced
the world's population. A new type of robots called Mecha, advanced humanoids
capable of thoughts and emotions, have been created.
No noo don’t stop to reading of my article, I
am saying this movie come to true in the late 22nd century.The goal of AI is to
provide software that can reason on input and explain on output. AI will provide human-like interactions
with software and offer decision support for specific tasks, but it’s not a
replacement for humans – and won’t be anytime soon. It augments our abilities
and makes us better at what we do. Because AI
algorithms learn differently than humans, they look at things differently. They
can see relationships and patterns that escape us.
“Neural Networks give thinking, Machine learning is popular but
deep learning breakthroughs drive AI
boom” says Arindam Bhadra, eSecurity expert cum blogger.
Now we look at why AI
will become an integral part of your video surveillance solution in the coming
years.
AI would play a major part in is going to be the analysis of
the massive amount of data that is collected by the security system. If you
consider the amount of data that is being collected by the device sensors, it
is difficult to make heads or tails of data at present and this is where AI
will come into the picture.
The rise of AI is
beginning to revolutionize technology in various industries, and security is
prominent among them.AI has become
more popular today thanks to increased data volumes, advanced algorithms, and
improvements in computing power and storage. Video surveillance solutions, for
instance, are now seeing an AI-based applications.
The AI CCTV is
basically a processing of the image with the help of deep learning algorithm
and after processing you will get information. Artificial intelligence CCTV
camera is just like CCTV operator that will give you information which you
mean.
I will give infographics comparison of 100 cases in which CCTV recording
as required, you can judge how to use artificial intelligence camera is.
To understand in the better way I will put an example so that
this topic will clear in the more better way.
If two camera is installed at the main gate of an industry
then what is the use of that camera, you need information and activity which is
happening at gate, like how many people entered in the industry, number of
vehicle passing from gate, may be you need vehicle number of all vehicle which
are entered in the premises, you may wish for how many visitor visiting in
premises, you may wish that how many time you regular staff passing from gate.
So these are the activities that you need to watch, for this you put CCTV
operator who can process all information and send you detail log but with help
of AI CCTV camera with AI-Software, you can find the automatic
log of these activities from software. So from the above example, it is more
clear that what actually artificial intelligence camera is.
Surveillance systems that include video analytics analyze
video footage in real-time and detect abnormal activities that could pose a
threat to an organization’s security. AI
helps security software ‘learn’ what is normal so it can identify unusual, and
potentially harmful, behavior that a human alone may miss.
To understand technically AI-CCTV camera is a simple IP camera but it is embedded with AI-Software, so according to need they
develop AI-software’s. There are
many kind algorithms and software likeFacial reorganization- it will allow you
finds people in a crowds place like railway stations, Airports, stadiums.
Intelligent Object, Event and Facial Recognition
In line with the concept of proactive and real-time security,
AI-based technologies can offer superior facial and object detection
capabilities. In fact, there is massive potential for facial recognition in the
law and order sector where specific individuals need to be identified from
crowds in large spaces.
And security is not the only place that is betting on this. Marketing departments in large corporations are also considering (if not already employing) facial recognition solutions to identify potential high-value customers who can be offered special services.
The power of AI will also enable “faceless recognition”, wherein a person can be identified not just from visuals of his face, but also from factors such as their height, posture, build, etc. Moreover, solutions could become more aware of the environment that they are monitoring, detecting patterns of activities and drawing out anomalies when required. They would also be able to experiment with different simulations and come up with possible scenarios where there could be potential threats.Faceless recognition systems would prove invaluable in situations such as the 2016 Berlin terrorist attacks where the attacker was caught on camera various times from angles where traditional biometric facial recognition systems failed.
And security is not the only place that is betting on this. Marketing departments in large corporations are also considering (if not already employing) facial recognition solutions to identify potential high-value customers who can be offered special services.
The power of AI will also enable “faceless recognition”, wherein a person can be identified not just from visuals of his face, but also from factors such as their height, posture, build, etc. Moreover, solutions could become more aware of the environment that they are monitoring, detecting patterns of activities and drawing out anomalies when required. They would also be able to experiment with different simulations and come up with possible scenarios where there could be potential threats.Faceless recognition systems would prove invaluable in situations such as the 2016 Berlin terrorist attacks where the attacker was caught on camera various times from angles where traditional biometric facial recognition systems failed.
Image Sharpening
Another huge benefit to using artificial intelligence in
video surveillance systems is its ability to cross reference data with other
neural networks or databases in order to fill the gaps in its own knowledge.
The first artificial network enhance the details of the image in accordance to
the layout provided in order to make it more feasible as an image.
This kind of image enhancing technology obviously has a huge
amount of potential in video surveillance and security, from criminal and
terrorist identification and tracking to staff monitoring and security access
operations. This kind of technology could also allow remote IP cameras to
stream enhanced video across a network that could then, in the case of a crime
scene investigation, be remotely accessed by a forensics expert to assist in
the investigation. The same could be true for security experts in video
surveillance use cases whereby detectives can access image-enhanced video to
investigate potential video recordings of wanted suspects.
Managing Massive deeper Data:
AI analyzes more and deeper data using neural networks
that have many hidden layers. Building a fraud detection system with five
hidden layers was almost impossible a few years ago. Surveillance is not just
about data from a single source anymore. Cameras and sensors of diverse types
are covering every nook and corner of locations to leave no room for error.
This means that security solutions must deal with substantial amounts of data
that require specialized software systems to manage.
AI systems could interpret data coming in via IP cameras and use it to automatically trigger alarms or notify security guards if it noticed something suspicious or out of place in the video data. Archive video footage could also be analyzed by AI systems using deep learning algorithms in order to place potentially new suspects at the scene of a crime automatically, with human agents merely updating a database the AI system had access to.
AI systems could interpret data coming in via IP cameras and use it to automatically trigger alarms or notify security guards if it noticed something suspicious or out of place in the video data. Archive video footage could also be analyzed by AI systems using deep learning algorithms in order to place potentially new suspects at the scene of a crime automatically, with human agents merely updating a database the AI system had access to.
Video Summary
Video Summary is a fast and powerful application that reduces
a long, archived video into a short, manageable video summary. Video Summary
reduces the time required to analyze archived video by removing unnecessary
video data such as static or empty scenes. Video Summary quickly analyzes the
video and removes the static background data while retaining the relevant data.
This allows security professionals to quickly review hours of archived video in
minutes. Video Summary can receive video input from many different types of video
archives recorded on DVRs, NVRs, or video management systems. It can also
provide real-time video summaries on live video feeds from analog, digital, or
IP/network cameras. With COM-SUR AI application help to Increase productivity
and efficiency of security professionals.
Alarm Center
Alarm Center provides options to view, search, report and
analyse all the alarms generated by AI analytics running on different systems
across a LAN. It also provides heat map for multiple cameras and provides the
paths of object or people movement. Reporting in PDF, JPG, Excel, Text file.
Unusual Motion Detection
UMD is an advanced artificial intelligence (AI) technology
that brings a new level of automation to surveillance and is designed to reveal
events that may have otherwise been missed. The analytic cameras will
support either object detection or UMD but, not both at the same time.
The analytics definitely work and would be used for perimeter
detection and alarming for sure but, if you aren't going to program the analytics
and be proactive about it then UMD could be turned on and the camera will learn
it's scene over the course of 1-2 weeks then when reviewing video instead of
looking at lots of different motion events on the timeline you can filter the
timeline to show just unusual events based on direction, speed, location.
"Today, most recorded video footage is never viewed,
meaning critical events may go undetected. Avigilon is using artificial
intelligence technology like UMD to help our customers effectively extract
valuable insights from their video data," said Dr. Mahesh
Saptharishi, Avigilon's CTO.
Edge Analytics
This is very cost effective solution without server
requirement. Applying some supported cameras can provide Intrusion (Trespass,
Boundary Crossing etc), Loitering, Crowding& Counting.
Enhanced Surveillance Monitoring
AI provide rapid detection of smoke even when smoke is formed
in 10-15% of the viewfor both indoor and Outdoor Environments. It’s called Video
based Smoke Detection(VSD).
Privacy Masking
Face and Skin area can be masked to protect the identity of
the persons appearing in the camera view of surveillance system. While unmasked
video is recorded & stored, masked video is viewed by operators during
surveillance monitoring.
Thermal Sensing Camera
Thermal camera will allow you to measure the temperature of
the particular object. It will install at a temperature sensitive place like
industrial boiler etc. accordingly analysis object details through AI software.
Trigger Based Detection
The trigger can be obtained from devices like RFID, Traffic
Light Signal. Triggers analytics for unauthorized people detection or object
detection after people leave from the area of interest. So you can say this is
apply for integration (BMS, ACS, Parking etc)
Real-Time Protection
Video surveillance solutions have always remained reactive
and useful only after an incident has occurred or at least a threat is detected
by human operators. This is all set to change with the arrival of AI systems
that can detect a potential threat way before an incident occurs.
For instance,deep learning, neural network can detect unattended objects like bags in airports and public spaces, enabling authorities to act before it’s too late. Similarly, AI-enabled surveillance solutions can monitor the behavior of people in a store and identify those who might engage in shoplifting.
For instance,deep learning, neural network can detect unattended objects like bags in airports and public spaces, enabling authorities to act before it’s too late. Similarly, AI-enabled surveillance solutions can monitor the behavior of people in a store and identify those who might engage in shoplifting.
Demographic Analysis
It detects the gender for human face appearing in the camera
field of view. Sometime is called Gender Detection.
This analysis can provide approximate age against human face
appearing in the camera field of view.
Parking Management
Provides the parking occupancy level and free parking slot
availability by detecting vehicle entry and exit by monitoring these transition
points in parking lot areas.
Also detects the percentage of area occupied by vehicles and
alerts against vehicle congestion as the vehicles occupy area beyond a
threshold value.Illegal parking (no parking zone or restricted zone) of
the vehicle in front of the entry/exit gates.
Beverage Management
Counts the number of cups / beverages passed over a line like
in serving counter area.
Queue Management
Detects queue properties for waiting time analysis (Ingress –
in flow, Egress – out flow. It’s applicable in service counters, ticket
counters / travel desk, check in counters, flight boarding areas for waiting
time analysis and improvement of service process.
Gesture Recognition
Detection of unusual gestures of persons by raising weapons
(lathi/stick, rifle etc.) with intention to charge on others. Or a person
slipping and falling on ground.Detection of object(s) removed / missing from
the monitored zone in the camera view.
Appearance Search
Appearance Search technology will allow users to initiate a
search for a person by selecting certain specific physical descriptions,
including hair and clothing color, gender, and age. The new feature is designed
to provide security operators enhanced speed and flexibility when conducting a
search for a person. "While investigating critical events, time is of the
essence, and security operators are often provided physical descriptions of the
people involved," said Dr. Mahesh Saptharishi, Avigilon's CTO.
Crowding Detection
This technology allows you to know the density of the object
in the crowd as the number of the object put at a particular place within the
camera field of view / region of interest.
AI Analyses crowd movement patterns / direction in different
way and marks it by different colours, and detects and movement of crowd in
undesired direction.
Measures/counting the crowd level in terms of number of
people occupying a specified region of interest in the camera field of view,
provides live crowd count on screen.
PTZ Analytics
Speed dome camera / scan dome camera is called PTZ camera. Basically
P – Pan, T – Tilt & Z indicate Zoom. Automatic tracking of object (Single
or Multiple) using Pan-Tilt-Zoom (PTZ) camera.
Violation detected on any Fixed camera triggers PTZ camera to
its view for auto tracking of the violator object.
Automatic one step PTZ operation to capture closer video of
an agent on any pre-set rule violation such as crossing a virtual line, or
entering a virtual area etc.
Different PTZ pre-set positions can be assigned to different
regions and analytics can be run to monitor each of those pre-set positions.
Camera Tampering
Detection of camera tampering efforts by camera focus change
or view obstruction or video cable cut.
ANPR
Number Plate Recognition Cameras allow you to recognize the
number of vehicle and along with that it also stored the vehicle log. This
camera is used in traffic management and traffic monitoring. It is also called
automatic number plate reorganization (ANPR).
Most public spaces in India including railway stations, bus
depots, metro stationand marketplaces are often under CCTV surveillance. New
Delhi is all set to have one of the largest deployments in the country of CCTVs
with the state government announcing plans to install 1.4 lakh CCTVs across
Delhi. The India Railways is also setting aside Rs 3,000 crore in its 2018-19
budget to install CCTV systems across 11,000 trains and 8,500 stations,
according to a news report.
“AI performs frequent, high-volume, computerized tasks
reliably and without fatigue” says Mr. Arindam Das Sarkar, Country Manager,
Mirasys India. Mirasys is one of the leader in creating models using deep
learning and Artificial intelligence based video analytics in India as well as
global market.
Tamper Detection
Quickly get security officers to the damaged camera in areas
with high crime rates, where there is a large potential for criminals purposely
blocking the camera’s view.
Additionally, several technologies enable and support AI:
·
Graphical
processing units are key to AI because they provide the heavy compute power
that’s required for iterative processing. Training neural networks requires big
data plus compute power.
·
The Internet of
Things generates massive amounts of data from connected devices, most
of it unanalyzed. Automating models with AI will allow us to use more of it.
·
Advanced
algorithms are being developed and combined in new ways to analyze more
data faster and at multiple levels. This intelligent processing is key to
identifying and predicting rare events, understanding complex systems and
optimizing unique scenarios.
·
APIs, or
application processing interfaces, are
portable packages of code that make
it possible to add AI functionality to existing products and software packages.
They can add image recognition capabilities to home security systems and
Q&A capabilities that describe data, create captions and headlines, or call
out interesting patterns and insights in data.
In Aug 2015, West Japan Railway Companymade a great example.
After discovering that 60 percent of people hit by trains in Japan are
intoxicated, they installed 46 security cameras that automatically
search for and detect signs of intoxication. The AIvideo looks for people who
are napping on benches, stumbling, falling, or standing
motionless for long periods of time and notifies human attendees who can then check
to see if the person is in danger of harming themselves.
Using AI in video
analytics, a number of systems will be able to communicate with each helping in
taking decisions and readily catching suspicious activities or predicting them
before they can happen.
With the advancement of AI,
the future of video analytics is not limited to providing stagnant algorithms,
but also to having the event and authorization-based alert systems. Wherein,
the alerts will be sent only to an authentic person and the relevant
department.