![]() |
You
can follow visitors´ route at the exhibition hall. |
There are four video cameras attached to the ceiling. Their view covers the whole hall. A computer programme linked to the cameras picks up people walking in the hall from the video pictures. On the monitor, you can see their movement as moving clusters of dots and quadrangles. The clusters leave a trail behind them. You can test the system by walking in the hall and checking your trail from the monitor. You can also change the monitor to “activity chart” mode, in which you’ll be able to see how people have been moving in the exhibition since its opening.
The machine vision system consists of cameras, computers and a programme that interprets the pictures. The characters of people are not standard in shape or colour so they can be localised only by movement. This means, that if a person stays put for a long time he/she will disappear from the system.
The lighting in the hall is in constant change during the day. This presents a challenge to the localisation system: neither the objects under surveillance nor the background stays standard in colour. Natural light is a problem for machine vision systems since it changes the appearance of objects and creates shadows. Whenever possible, machine vision systems use fixed lighting.
The system has been programmed to neglect some objects that are in motion in the hall all the time. For example the Hot air balloon and the Foucault’s pendulum are such objects. Also, the colour of a person’s clothing can make localisation difficult if it resembles the shade of the floor.
Machine vision is being used in traffic surveillance as well as in quality control. For example, the quality of products on an industrial conveyer belt can be monitored. Machine vision is suitable for tasks which require quick, accurate, repeatedly performed, 24/7 optical examination.
The pictures that the cameras take of the hall are not being saved anywhere. They are visible only on the adjoining monitor.
Realisation of the
programme
Sami Laakso, Tommi Tykkälä
Updated 8.1.2007/KT

