Cobalt detection against complex foregrounds

People moving in the smoke of a fire (Image-DNA example)

This 2 minute video shows a scene on a building site where rubbish is being burned on a smoky fire. The swirling smoke covers the entire scene, providing a foreground behind which movements must be detected.

The scene in this 2 minute video is dominated by the swirling smoke from the fire in the center of the building site. The smoke should not produce alarms while other activity is detected. While people move around the site they are indeed detected, as revealed by the flashing yellow boxes that their movement triggers. There is no audio.

Setting up these detection zones merely required that the operator draw a few boxes: there were no parameters to set and no other decisions to make. The simplicity of the setup of course means that anything moving across the field of view, in any direction, would be detected.

The Cobalt software was specifically targeted at handling the problem of complex ever-changing foregrounds and backgrounds. The key to making this work was the use of Image-DNA which, among other things, characterizes textures and can distinguish changes in texture. Here the background sea has a particular texture that is quite different from that of the seagulls.