SMART SURVEILLANCE

Presentation by:Puneet Soni

Saurabh BhatiaPurvesh DwivediPatil Abhishek

What does SMART SURVEILLANCE means?????

The proposal is to use the concepts of artificial intelligence and digital image processing in

surveillance systems to detect crime or crime related events, threats and notify officials accordingly. The system could monitor the visual and audio data obtained from the

security cameras in-built with microphones and process them to detect crime or crime related

events and notify the officials accordingly.

“Technologies Used in SMART SURVEILLACE”

Plug-and-play analytics frameworksObject detection and trackingAlert definition and detectionObject and colour classificationDatabase event indexingSearch and retrieval

Plug-and-play analytics frameworks

Video cameras capture a wide range of information about people, vehicles, and events. The type of information

captured is dependent on a number of parameters like camera type, angle,

field of view, and resolution. Automatically detecting each

type of information requires specialized sets of algorithms.

Object detection and tracking

One of the core capabilities of smart surveillance systems is the ability to

detect and track moving objects. Object detection algorithms are typically statistical learning algorithms that

dynamically learn the scene background model and use the reference model to

determine which parts of the scene correspond to moving objects.

Alert definition and detection

Typical smart surveillance systems support a variety of user-defined behaviour detection capabilities such as detecting motion within a defined zone, detecting objects that cross

a user defined virtual boundary, and detecting objects that are abandoned.

Graphical user interface (GUI) tools are used to define zones of interest, object sizes, and

other parameters needed to define the behaviour.

Object and colour classification

Object classification algorithms classify objects into different classes, for

example, People, Vehicles, Animals, and use training data and calibration

schemes. Colour classification classifies the

dominant colour of the object into one of the standard colours (red, green,

blue, yellow, black, and white).

Database event indexing

The events detected by the video analysis algorithms are indexed by content and stored in

a database. This allows events to be cross-referenced across multiple spatially distributed

cameras and creates a historical archive of events. The event index information typically

includes time of occurrence, camera identifier, event type, object type, object appearance

attributes, and an index into the video repository which allows the user to “play back the relevant

video at the touch of a button.”

Search and retrieval

Users can use a variety of GUI tools to define complex search criteria to retrieve specific events. Search criteria include,

object size, colour, location in the scene, velocity, time of occurrence, and several

other parameters.The results of a search can also be

rendered in a variety of summary views.

Implications of Smart Surveillance

Smart surveillance is a technology that has many different applications and potentially has significant implications to each of these. We look at implications primarily in the surveillance application, namely, security and privacy.

Security Implications: Clearly, the ability to provide real time alerts, capture high value video and provide sophisticated forensic video retrieval has the potential to enhance security in various public and private facilities.

IBM Smart Surveillance System

References

An article on Artificial Intelligence http://www.formal.stanford.edu/jmc/whatisai/whatisai.html

An article about research involving use of IT for visual processing in surveillance systems http://www.phys.org/news/-10-surveillance-tech-carnegie-mellon.html

A Oltramari, “Using Ontologies in a Cognitive-Grounded System”, 2012 STIDS

An article on use of acoustics in forensics http://www.acousticassociates.com/ForensicAcousticsSummary.aspx

Russell, Stuart J.; Norvig, Peter (2003), Artificial Intelligence: A Modern Approach (2nd ed.), Upper Saddle River, New Jersey