next up previous
Next: Coping with Real Data Up: Capturing Dense Environmental Range Previous: Hardware Teaching Lab

Tracked Acquisition

An additional application for our rangefinder is data collection for multiple-center-of-projection (MCOP) images. Think of these as images where each column of data is acquired while the rangefinder is moving. If the position and orientation of the rangefinder is well known for each sample (or group of samples), then the scene can be properly reconstructed. This style of image allows close-up data collection where detail is necessary and quick data collection from afar when the data is not so important. See [3] for details on MCOP images.

To accommodate the MCOP project, we mount a hi-ball tracking device [1] on the rangefinder platform (shown in figure 5 on the left). The hi-ball is a device with 6 lenses looking towards a ceiling with sequentially flashing IR LEDs. The sensing of the flashes provides highly accurate position and orientation data.

The data collection proceeds by collecting each column with its position and orientation. Examples of the data collected and their reconstruction are shown in figures 16 and 17. This example shows that not all input must be in regular images, but can come from a variety of sampling techniques.


  
Figure 16: Data collected while the scanning rangefinder was slowly pushed across the room. The position of the rangefinder is recorded when each column of data is collected. Note the irregular motion in the distorted ceiling tiles, walls and checkerboard patterns.
\begin{figure}\begin{center}
\epsfig{file=phil_color.eps,width=21pc}
\end{center}\end{figure}


  
Figure 17: Reprojection of the data collected in figure 16. Note how the ceiling tiles and checkerboard patterns are regular and the walls, tables and ceiling are flat.
\begin{figure}\begin{center}
\epsfig{file=phil-reproject1.eps,width=21pc}
\end{center}\end{figure}


next up previous
Next: Coping with Real Data Up: Capturing Dense Environmental Range Previous: Hardware Teaching Lab
Lars S. Nyland
1999-02-19