Keeping an eye on suspicious behaviour

Siemens has developed a video surveillance system known as Sistore that automatically identifies intruders and tracks them via a series of cameras.

Sistore digitises the images recorded by video cameras in real-time, and then uses a special algorithm to search them for suspicious movements. If it discovers an intruder, the system will then trigger an alarm in the security control centre, and automatically begin tracking the person in question with other available cameras.

Sistore prevents security guards having to keep their eyes on several screens at once, which can quickly cause fatigue. Studies have shown that even when watching only two monitors with automatic camera switching, a guard will miss nearly half of all activity in a scene after 12 minutes and 95 per cent after 22 minutes.

The video sensors in Sistore ‘learn’ what the normal situation in a camera’s image should be by memorising the most frequent conditions within a specified period of time. They then ignore an unchanging environment and focus on any deviations.

The system works with algorithms for attribute extraction, where adjustable parameters such as size and speed enable it to distinguish between a person and a car or animal.

The memorised normal background image also enables the system to automatically recognise sabotage, so if someone turns the camera around, the background image will change, and the system will sound an alarm.

Sistore can be combined with a customer’s existing intranet structure, which means any number of video signals can be linked to it and analysed. Because Sistore uses the Internet as a system platform, security guards can view images on a PC monitor from theoretically anywhere in the world.

Siemens is installing a fully digitised Sistore system for the 2006 Asian Games in Doha. Some 1,300 cameras will be linked up to the system to help keep the capital of Qatar safe during the games, which will take place in the first two weeks of December.