Mobile Robot Navigation using Ground Plane Detection^(a)

   Under Construction...                                                                          by Daniel Conrad

 Introduction

    The idea behind this research is that when it comes to vision-based mobile robotic navigation, one of the most important features to locate in a scene is the ground plane. Once this plane is found, all other objects in a scene can be segmented out. To detect the ground plane, we propose a homography based approach that uses a Modified Expectation Maximization (MEM) algorithm.
We tested our approach by creating a simple target following algorithm. This algorithm used our MEM approach to detect the ground plane in the scene. Once the ground plane was found, the rest of the feature points we assumed to belong to the target that we are suppose to follow. Results can be seen in the following sequences:

    References

1.  Conrad, D. and DeSouza, G. N., "Homography-based Ground Plane Detection for Mobile Robot Navigation Using a Modified EM Algorithm". IEEE 2010 International Conference on Robotics and Automation (ICRA), pp.910-915, May/2010, Anchorage, Alaska, USA.
    
2. DeSouza G.N., Kak A.C., " Vision for Mobile Robot Navigation", IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 24, No. 2, pp. , Feb. 2002.

^(a) This research made use of GTX480s and Tesla's S1070 donated by NVIDIA via their Academic Partnership Program.