Page - 172 - in Proceedings of the OAGM&ARW Joint Workshop - Vision, Automation and Robotics

[8] (25Hz) additionally implement some kind of obstacle detection but do not utilize pre-filtering. Apart from this, the conditions are reasonably similar. On the hardware side all results where achieved on an Intel Core i7 with around 2GHz while utilizing only one CPU core and the GPU for pre filtering. VII. CONCLUSION We introduced a new plane segmentation approach for 2.5D data. It shows competitive results for both, quality and speed. Our algorithm relies on a filtering step that improves the quality of the input data. Hence, we conducted an analysis of three filters to find a fitting candidate. We selected the Sigma Adaptive Bilateral Filter wich bal- ances speed and quality. Our GPU implementation of the filter algorithm runs within 25ms on an AMD Radeon HD 6750M. The mesh denoising algorithm [4], together with our modifications showed promising results. To utilize this algorithm in real-time, GPUs with higher performance could be a possible solution. Apart from this, both filters could be improved by adding a noise model that handles the increased noise levels at higher distances. The proposed segmentation algorithm shows competitive re- sults that were achieved with a hierarchical strategy. Splitting up thesegmentation intoacoarsepre-segmentationandafine grained post-processing step holds the run-time competitive. Future work could extend this algorithm with a sensor model that leads to additional rules such as e.g. depth dependent thresholds. REFERENCES [1] K. Bredies, K. Kunisch, and T. Pock, “Total generalized variation,” SIAM J. Img. Sci., vol. 3, no. 3, pp. 492–526, Sep. 2010. [2] A. Deutschmann,KantenselektiverFilter fu¨rPunktwolken. TU Wien, 2013. [3] M. A. Fischler and R. C. Bolles, “Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography,” Commun. ACM, vol. 24, no. 6, pp. 381–395, Jun. 1981. [Online]. Available: http://doi.acm.org/10.1145/358669.358692 [4] S. Fleishman, I. Drori, and D. Cohen-Or, “Bilateral mesh denoising,” inACMSIGGRAPH2003 Papers, ser. SIGGRAPH ’03. New York, NY, USA: ACM, 2003, pp. 950–953. [5] D. Holz, S. Holzer, R. B. Rusu, and S. Behnke, Real-Time Plane Segmentation Using RGB-DCameras. Berlin, Heidelberg: Springer Berlin Heidelberg, 2012, pp. 306–317. [6] L. I. Rudin, S. Osher, and E. Fatemi, “Nonlinear total variation based noise removal algorithms,”PhysicaD:NonlinearPhenomena, vol. 60, no. 1, pp. 259 – 268, 1992. [7] C. Tomasi and R. Manduchi, “Bilateral filtering for gray and color images,” inSixth InternationalConference onComputerVision (IEEE Cat. No.98CH36271), Jan 1998, pp. 839–846. [8] Z. Wang, H. Liu, Y. Qian, and T. Xu,Real-Time Plane Segmentation andObstacleDetectionof3DPointClouds for IndoorScenes. Berlin, Heidelberg: Springer Berlin Heidelberg, 2012, pp. 22–31. [9] D. Yiruo, W. Wenjia, and K. Yukihiro, “Complex ground plane detection based on v-disparity map in off-road environment,” in 2013 IEEE Intelligent Vehicles Symposium (IV), June 2013, pp. 1137–1142. [10] L. Zhang, D. Chen, and W. Liu, “Fast plane segmentation with line primitives for rgb-d sensor,” International Journal of Advanced Robotic Systems, vol. 13, no. 6, p. 8, 2016. [11] J. Zhao, J. Katupitiya, and J. Ward, “Global correlation based ground plane estimation using v-disparity image,” inProceedings 2007 IEEE InternationalConferenceonRoboticsandAutomation, April 2007, pp. 529–534. 172