The Secrets of Security Video Analytics

The Secrets of Security Video Analytics

This post is also available in: heעברית (Hebrew)

Due to improvements in the field, the demands of connected cameras today are increasing.

One of the technologies that security cameras offer us today is Video Motion Detection – or VMD. The purpose of this technology is to make sure the cameras detects the move perpetrator in the “protected zone”, thus triggering an external device (such as alarms, text messages, emails and such).

The way this (outdated) technology works is this: The camera “remembers” an existing frame when it is set up. While detecting a change in a certain percent of the frame’s pixels – the cameras starts recording to a remote VDR unit. This causes two main problems: The first is false detection. The camera can’t tell when the objects moving in the frame are animals, birds, clouds, branches and small objects, and so the client might reject the product. The second problem derives from the first and is “clogging” the storage sources (the recording unit’s hard-drive), since the camera records almost constantly.

But today there is growing use of analytics technology, meaning the smart analysis of the frame. The use of analytics offers to distinguish true alarms from different movements happening in the camera’s sight, thus reducing instances of false alarms to nearly zero. The design of analytics is unique to every sight and every purpose. It begins with the reviewer’s demand and ends with field adjustments by the installer.

The possibilities of this product are many and diverse, and include establishing a sterile zone where every movement of a person entails an immediate alarm, establishing a zone where every prolonged movement (for a pre-defined period of time) in a defined zone, movement at a defined direction, detecting sabotage, establishing a manned zone, and more.

Avigilon’s analytics software (formerly Video-IQ) offers to quickly pull data in retrospect to investigate an event, even if no analytical condition was set in the system. Meaning you can easily trace a forklift theft, for example, from a zone which was not defined as sterile. The system can identify it on its own, turn to the exact time and present the incident in question. The system operates in outdoors conditions so that movement of birds, clouds, plants, etc. cause no false alarms.

6What makes Avigilon’s analytics system so unique is its ability to independently learn the field conditions and to calibrate itself as to provide more accurate results with time. It also allows the user to help it learn by stating whether the  incidents were true or false. When the user points to a mistake the system made, it learns from it never to repeat it. This way of learning is similar to the way we humans learn in all aspects of our lives.

The system was designed by Avigilon after the company saw the need for easy-to-operate yet accurate analytics solutions. The frame analysis is done by designated compressors or by the camera itself. Even in the case of intentional sabotage, the camera will keep a local recording. Even in the case of failure in communication lines, events are still stored inside the camera. When the lines return to operate, the camera will send the images to the storage unit.

Avigilon’s official website.

Written by Itamar Kerbel – Technical Manager at Polar Electronic Systems ltd.