Why metadata is better data for video analytics!

Justin Schorn is VP of Product Management and co-founder of Aimetis.

Aimetis Corporation uses metadata technology to create cutting edge products for physical security, surveillance and business intelligence applications.

We asked Justin Schorn, VP of Product Management and co-founder of Aimetis, to outline how metadata works and explain how it can be used to help protect and manage sites.

What is metadata?

“Metadata” is a term used to describe data that informs us about other data. In surveillance terms, it is the data relating to changes in a video stream. Metadata provides identity and context to events in the digital video stream, allowing the video to be organised, searched and retrieved easily.

“Machine understandable”

Digital video content is inherently unstructured and difficult to search. Machines cannot “watch” video and interpret it like a human. However, metadata uses identifiers to inform a computer what it is “watching”. Identifiers are absolute, measureable and scalable facts or events. They can include time, location and other application-specific attributes. They enable a machine to analyse content for changes and trends and search for specific types of event. Metadata makes video data “machine-understandable”.

Metadata and security

Metadata offers three significant security surveillance functions - search, alert/alarm and reporting. Metadata adds context to each digital video frame and the objects, vehicles and people within it, including:

  • Date and time
  • Camera obstructions
  • Scene motion
  • Camera movement

Metadata allows digital video information to be far more “searchable” by defining:

  • Object identity
  • Object size
  • Object time in view
  • Object type e.g. car, person, train etc
  • Object movement
  • Object size in a scene
  • Object location in a frame
  • Vehicle license plate numbers

Security management functions use these specific triggers to identify incidents and events. Rules can be set up, such as ‘virtual fences’, trip wires and people counting, to help understand the interactions objects have with their environment. A scene can be learnt over a period of time and normal object movement (or lack of movement) tracked within that scene. “Normal” or “abnormal” events can be recognised against a defined set of rules, with only the changes in a scene being registered, not the entire scene.

Metadata and multiple camera streams – where to save money!

Many leading camera manufacturers provide multiple video streams, simultaneously, from a single camera. The Full HD stream is usually used for high-quality recording and the other, lower resolution streams are optimised for viewing by PC’s, smartphones and tablets. Aimetis software will record the Full HD stream in this manner, but can save processor capability by using one of the smaller video streams to generate metadata for analytics. This is possible because the metadata relates to the “changes” occurring in a video stream, which are often the same in a 1080p, 720p, D1 or CIF stream.

Producing metadata from a low resolution stream and using it to identify and track events on a Full HD stream is a highly efficient approach to analytics. It is a hugely important breakthrough in IP CCTV system management and it is only possible when the VMS and analytics are integrated within the same software package, like Aimetis. In practical terms, it means less processing power is needed to achieve the same levels of VMS and analytics capability. Reduced processing demands mean more video channels can be processed per server, reducing the cost of the system.

Analytics functions and system management

Reports about specific object behaviour can be developed, allowing security management to become very efficient and knowledgeable. Metadata makes it possible to track and understand object behaviours like never before. Alerts for missing object, moved object, and change of object can be set; a vital function for any art gallery, stately home or museum with objects on display. New objects introduced to a scene can be identified and trigger an alarm in busy areas, such as airports, railway stations and town centres. Human behaviour can be learnt and tracked so entering, loitering or exiting a building or zone can be monitored.

Multiple events and incidents

Metadata allows very complex activity to be analysed, where security operators would struggle to watch and identify every significant aspect of a changing scene. Multiple events, such as at large public gatherings and areas of heavy use, can be analysed simultaneously. This leads to early, real-time detection of abnormal and possibly suspicious behaviour, enabling crime prevention and interception in security applications.

Retail management

Metadata allows details of store, vehicle and visitor traffic and behaviour to be collated. This includes footfall mapping, people counting, entrance and exit activity, dwell times, peak times, car park behaviour, delivery vehicle monitoring and till activity. It can dramatically influence efficiency, providing information for security management, staffing levels and staff deployment, car park management, deliveries, store layout, marketing and logistics.

Metadata can assist in remote and proactive security management. Notifications and alerts can be sent to any required device, integrating surveillance with mobile technologies and ensuring the right person receives the relevant information immediately. Metadata technology provides site management and business enhancement functionality to those responsible for large sites, such as airports, ports, car parks and hospitals.