A CPU-GPU Hybrid System of Environment Perception and 3D Terrain Reconstruction for Unmanned Ground Vehicle
0
조회
0
다운로드
3
인용
0
참고문헌
키워드
초록
Environment perception and three-dimensional (3D) reconstruction tasks are used to provide unmannedground vehicle (UGV) with driving awareness interfaces. The speed of obstacle segmentation and surroundingterrain reconstruction crucially influences decision making in UGVs. To increase the processing speed ofenvironment information analysis, we develop a CPU-GPU hybrid system of automatic environmentperception and 3D terrain reconstruction based on the integration of multiple sensors. The system consists ofthree functional modules, namely, multi-sensor data collection and pre-processing, environment perception,and 3D reconstruction. To integrate individual datasets collected from different sensors, the pre-processingfunction registers the sensed LiDAR (light detection and ranging) point clouds, video sequences, and motioninformation into a global terrain model after filtering redundant and noise data according to the redundancyremoval principle. In the environment perception module, the registered discrete points are clustered intoground surface and individual objects by using a ground segmentation method and a connected componentlabeling algorithm. The estimated ground surface and non-ground objects indicate the terrain to be traversedand obstacles in the environment, thus creating driving awareness. The 3D reconstruction module calibratesthe projection matrix between the mounted LiDAR and cameras to map the local point clouds onto thecaptured video images. Texture meshes and color particle models are used to reconstruct the ground surfaceand objects of the 3D terrain model, respectively. To accelerate the proposed system, we apply the GPU parallelcomputation method to implement the applied computer graphics and image processing algorithms in parallel.
Abstract
Environment perception and three-dimensional (3D) reconstruction tasks are used to provide unmannedground vehicle (UGV) with driving awareness interfaces. The speed of obstacle segmentation and surroundingterrain reconstruction crucially influences decision making in UGVs. To increase the processing speed ofenvironment information analysis, we develop a CPU-GPU hybrid system of automatic environmentperception and 3D terrain reconstruction based on the integration of multiple sensors. The system consists ofthree functional modules, namely, multi-sensor data collection and pre-processing, environment perception,and 3D reconstruction. To integrate individual datasets collected from different sensors, the pre-processingfunction registers the sensed LiDAR (light detection and ranging) point clouds, video sequences, and motioninformation into a global terrain model after filtering redundant and noise data according to the redundancyremoval principle. In the environment perception module, the registered discrete points are clustered intoground surface and individual objects by using a ground segmentation method and a connected componentlabeling algorithm. The estimated ground surface and non-ground objects indicate the terrain to be traversedand obstacles in the environment, thus creating driving awareness. The 3D reconstruction module calibratesthe projection matrix between the mounted LiDAR and cameras to map the local point clouds onto thecaptured video images. Texture meshes and color particle models are used to reconstruct the ground surfaceand objects of the 3D terrain model, respectively. To accelerate the proposed system, we apply the GPU parallelcomputation method to implement the applied computer graphics and image processing algorithms in parallel.
참고문헌 (16)
- Y. Matsushita, "On-line road boundary modeling with multiple sensory features, flexible road model, and particle filter", Robotics and Autonomous Systems, 59(5), pp.274~284 (2011)
- J. M. Noguera, "Navigating large terrains using commodity mobile devices", Computers & Geosciences, 37(9), pp.1218~1233 (2011)
- W. Song, "Classifying 3D objects in LiDAR point clouds with a backpropagation neural network", Human-centric Computing and Information Sciences, 8 (2018)
- S. R. Sukumar, "Multi-sensor integration for unmanned terrain modeling", Proceedings of SPIE 6230: Unmanned Systems Technology VIII (2006), International Society for Optics and Photonics
- D. Huber, "Real-time photo-realistic visualization of 3D environments for enhanced tele-operation of vehicles", Proceedings of 2009 IEEE 12th International Conference on Computer Vision Workshops (ICCV Workshops), pp.1518~1525 (2009)
- J. Elseberg, "One billion points in the cloud–an octree for efficient processing of 3D laser scans", ISPRS Journal of Photogrammetry and Remote Sensing, 76, pp.76~88 (2013)
- D. Gingras, "Rough terrain reconstruction for rover motion planning", Proceedings of 2010 Canadian Conference on Computer and Robot Vision (CRV), pp.191~198 (2010)
- A. Golovinskiy, "Min-cut based segmentation of point clouds", Proceedings of 2009IEEE 12th International Conference on Computer Vision Workshops (ICCV Workshops), pp.39~46 (2009)
- B. Douillard, "Experimental Robotics", pp.585~600 (2014), Springer
- M. Himmelsbach, "Fast segmentation of 3D point clouds for ground vehicles", Proceedings of 2010 IEEE Intelligent Vehicles Symposium, pp.560~565 (2010)
- J. Wang, "SigVox: a 3D feature matching algorithm for automatic street object recognition in mobile laser scanning point clouds", ISPRS Journal of Photogrammetry and Remote Sensing, 128, pp.111~129 (2017)
- H. Wang, "Pedestrian recognition and tracking using 3D LiDAR for autonomous vehicle", Robotics and Autonomous Systems, 88, pp.71~78 (2017)
- A. Broggi, "A full-3D voxel-based dynamic obstacle detection for urban scenario using stereo vision", Proceedings of 2013 16th International IEEE Conference on Intelligent Transportation Systems-(ITSC), pp.71~76 (2013)
- Alireza Khatamian, "Survey on 3D Surface Reconstruction", JIPS(Journal of Information Processing Systems), 12(3), pp.338~357 (2016), 한국정보처리학회
- W. Song, "A 3D localisation method in indoor environments for virtual reality applications", Human-centric Computing and Information Sciences, 7 (2017)
- D. Zeng, "Adversarial learning for distant supervised relation extraction", Computers, Materials & Continua, 55(1), pp.121~136 (2018)