Camera placement for Visual Tagging

Jian Zhao, Sen-ching S. Cheung

Graphic User Interface

This Graphic User Interface takes a floor-map as an input and allows the user to define obstacles, candidate area for tags and camera placement as well as emphasized regions for surveillance. The camera and tag grids can be either generated according to the user input automatically or manually input by the user. The GUI also allows the user to input other parameters such as the number of cameras available and the minimum number of camera views at which that tag at each tag grid point must be visible.

After specifying all the required parameters, the GUI will feed the parameters to the optimization algorithms and the resulting camera configuration will be overlayed on top of the floor-plan. A simulation map will also be shown indicating the result of continuous space

.

3-D Simulation 1

This demo shows the 3-D computer simulation of single humanoid in an empty room. At each time instance, there are at least two cameras observing the red tag.

3D simulation 2

Locating by Epipolar lines

The upper two views come from two cameras with a tag detector. The video in the bottom shows how the third camera with the detector disabled can infer the tag's position using epipolar lines from the two cameras above.

The three cameras are placed as in the following figure

Locating by 3D triangulation

Using fully calibrated cameras, we can triangulate the tag centroid using the two observations and back-project it to the third camera, achieving similar result as the epipolar line.

Related Publications: