next up previous contents
Next: Conclusion Up: Some results Previous: Architecture and heritage conservation   Contents

Planetary rover control

In this section a system is presented that we developed for ESA for the support of planetary exploration. More information can be found in [167,168,169]. The system that is send to the planetary surface consists of a rover and lander. The lander has a stereo head equipped with a pan-tilt mechanism. This vision system is used both for modeling of the terrain and for localization of the rover. Both tasks are necessary for the navigation of the rover. Due to the stress that occurs during the flight a recalibration of the stereo vision system is required once it is deployed on the planet. Due to practical limitations it is infeasible to use a known calibration pattern for this purpose and therefore a new calibration procedure had to be developed that can work on images of the planetary environment. This automatic procedure recovers the relative orientation of the cameras and the pan- and tilt-axis, besides the exterior orientation for all the images. The same images are subsequently used to recover the 3D structure of the terrain. For this purpose a dense stereo matching algorithm is used that -after rectification- computes a disparity map. Finally, all the disparity maps are merged into a single digital terrain model. This procedure is illustrated in Figure 9.24.

Figure 9.24: Digital Elevation Map generation in detail. The figure contains the left rectified image (left), the corresponding disparity map (middle) and the right rectified image (right). The 3D line $L$ corresponding to a point of the DEM projects to $L_1$ resp. $L_2$. Through the disparity map the shadow of $L_1$ on the right image can also be computed. The intersection point of $L_2$ and $L_1$ corresponds to the point where $L$ intersects the surface.
\begin{figure}\centerline{\psfig{figure=results/,width=120mm}} \end{figure}
The fact that the same images can be used for both calibration and 3D reconstruction is important since in general the communication bandwidth is very limited. In addition to the use for navigation and path planning, the 3D model of the terrain is also used for Virtual Reality simulation of the mission, in which case the model is texture mapped with the original images. A first test of the complete system was performed at the ESA-ESTEC test facilities in Noordwijk (The Netherlands) where access to a planetary testbed of about 7 by 7 meters was available. The Imaging Head was set up next to the testbed. Its first task was the recording of the terrain. A mosaic of the pictures taken by this process can be seen in figure 9.25.
Figure 9.25: Mosaic of images of the testbed taken by the stereo head.
\begin{figure}\centerline{\psfig{figure=results/,width=150mm}} \end{figure}
The autonomous calibration procedure was launched and it computed the extrinsic calibration of the cameras based on the images. Once the calibration had been computed the system rectified the images and computed dense disparity maps. Based on these, a Digital Elevation Map was constructed. The result can be seen in figure 9.26.
Figure 9.26: Digital Elevation Map of the ESTEC planetary testbed. A significant amount of cells is not filled in because they are located in occluded areas.
\begin{figure}\centerline{\psfig{figure=results/,width=70mm}} \end{figure}
Because of the relatively low height of the Imaging Head (approximately 1.5 meters above the testbed) and the big rocks in the testbed, a large portion of the Digital Elevation Map could not be filled in because of occlusions.

It can be expected that global terrain model will not be sufficiently accurate to achieve fine steering of the rover for some specific operations. An important task of the rover is to carry out surface measurements on interesting rocks. For this purpose it is important to be able to very accurately position the measurement device. Since probably the rover will also be equipped with a (single) camera the technique proposed in this text could be used to reconstruct a local model of an interesting rock while the rover is approaching it. A preliminary test was carried out on a rock in the ESA testbed. Some results can be seen in Figure 9.27.

Figure 9.27: Different steps of the 3D reconstruction of a rock from images.

next up previous contents
Next: Conclusion Up: Some results Previous: Architecture and heritage conservation   Contents
Marc Pollefeys 2002-11-22