Terrain Classification Based On Structure For Autonomous Navigation in Complex Environments Duong V.Nguyen 1, Lars Kuhnert 2, Markus Ax 2, and Klaus-Dieter.

Terrain Classification Based On Structure ForAutonomous Navigation in Complex Environments

Duong V.Nguyen1, Lars Kuhnert2, Markus Ax2, and Klaus-Dieter Kuhnert2

1Research School MOSES, University of Siegen, Germany

2Institute for Real-Time-Learning Systems, University of Siegen, Germany

II. Signal Processing And Application

Introduction

Methodology• Graph-Cut

• Feature Extraction

• Neighbor Distance Variation Inside Edgeless Areas

• Conditional Local Point Statistics

• Support Vector Machine

Experiments and Results

Conclusion

Reference

Outline

Introduction

•Variety of terrain •Avoid obstacles• Maintain rollover stability• Manage power …etc

Why do we need Terrain Classification?

autonomous operation Or: complete task without direct control by a human• Bomb-defusing • Vacuum cleaning • Forest exploration …etc

What is unmanned system ?

AMOR:

1st prize of innovation awards, ELROB-2010, Hammelburg, Germany.

PMD camera Laser Scanner Stereo Cameras

Introduction Recent 3-D Approaches

Problems: Beam scattering effects Only used for static scenes Object detection purely based on structure is

not really robust in some scenes.

Solutions: Local points statistic analysis (Graph-Cut for depth image segmentation) Gaussian Mixture Model using Expectation

maximization Combining 3-D and 2-D features

Why should Laser Scanner be used?

Advantages: Stable data acquisition High precision Affordable

Introduction

Classifier SVM

ROI extraction

3-D point cloud

3-D Features Depth image segmentation

Methodology

Terrain Classification System Diagram

Graph-Cut Technique

Methodology

Internal difference

Component difference

Un-Joint Condition:

Classifier SVM

ROI extraction

3-D point cloud

3-D Features Depth image segmentation

Methodology Feature Extraction

Classifier SVM

ROI extraction

3-D point cloud

3-D Features Depth image segmentation

Methodology Support Vector Machine

Experiments and Results

• Graph-cut Technique For Segmentation• Neighbor Distance Variation Feature• Conditional Local Point Statistics Feature

Future work:• 2D&3D Calibration• Color Features

Conclusion

Q&A

References[1] David Bradley, Ranjith Unnikrishnan, and J. Andrew (Drew) Bagnell, Vegetation Detection for Driving in Complex Environments, IEEE International

Conference on Robotics and Automation, April, 2007.

[2] Matthias Plaue, Analysis of the PMD Imaging System, Technical Report,Berlin, Germany, 2006.

[3] Duong V.Nguyen, Lars Kuhnert, Markus Ax, and Klaus-Dieter Kuhnert,Combining distance and modulation information for detecting pedestrians in outdoor environment using a PMD camera, Proc. on the 11th IASTED International Conference Computer Graphics and Imaging, Innsbruck, Austria, Feb-2010.

[4] John Tuley, Nicolas Vandapel, and Martial Hebert, Technical report CMU-RI-TR-04-44, Robotics Institute, Carnegie Mellon University, August, 2004.

[5] J. F. Lalonde, N. Vandapel, D. F. Huber, M. Hebert, Natural Terrain Classification using Three-Dimensional Ladar Data for Ground Robot Mobility, Journal of Field Robotics, Volume 23 Issue 10, Pages 839 -861, Oct-2006.

[6] Pedro F. Felzenszwalb, Daniel P. Huttenlocher , Efficient Graph-Based Image Segmentation, IJCV, Volume 59 Issue 2, Sept-2004.

[7] R. Willstatter and A. Stoll, Utersuchungenuber Chlorophyll, Berlin: Springer, 1913.

[8] Anguelov, D., Taskar, B., Chatalashev, V., Koller, D., Gupta, D., Heitz, G. Ng, A.,Discriminative Learning of Markov Random Fields for Segmentation of 3D Scan Data, ICVPR, San Diego, CA, USA, 2004.

[9] Huang, J., Lee, A. Mumford, D. Statistics of Range Images, ICVPR, Los Alamitos, CA, USA, 2000.

[10] Rasmussen, C., Combining Laser Range, Color and Texture Cues for Autonomous Road Following, ICRA, Washington, DC, USA.

[11] N. Vandapel and M. Herbert, Natural terrain classification using 3-d ladar data, in IEEE Int. Conf. on Robotics and Automation (ICRA), 2004.

[12] C. Cortes, V. Vapnik, Support vector networks. Machine Learning, vol.20, p. 273-297, 1995.

[13] Quinlan, J. 1993. C4.5: Programs for Machine Learning. Morgan Kaufmann: San Mateo, CA.

[14] Guoqiang Peter Zhang, Neural Networks for Classification: A Survey, IEEE Trans. on Systems, Man, and Cybernetics, Vol. 30, No. 4, Nov-2000.

[15] Y. Freund and R. E. Schapire, A decision-theoretic generalization of on-line learning and an application to boosting, Journal of Computer and System Sciences, no. 55. 1997.

[16] Bilmes, J., A Gentle Tutorial on the EM Algorithm and its Application to Parameter Estimation for Gaussian Mixture and Hidden Markov Models, Berkeley, CA, USA: The International Computer Science Institute, University of Berkeley, Technical Report, 1997.

[17] Baudat. G and Anouar. F Kernel-Based Methods and Function Approximation, International Joint Conference on Neural Networks, USA, pp.

1244 - 1249, 2001.