Unsupervised Learning and Reverse Optical Flow in Mobile RoboticsStanford University, 2011 As sensor resolution increases and costs decrease, the amount of data available on mobile robotics platforms is exploding. Unsupervised machine learning algorithms, and their ability to produce useful information without large labeled training sets, are an important tool for benefiting from this abundance. In this thesis the application of unsupervised learning to three subfields of mobile robotics is discussed. Tracking multiple moving objects from an unmanned aerial vehicle, road following in loosely-structured environments, and autonomous offroad navigation. The thesis focuses on building dynamic activity-based ground models for multi-object tracking, the combination of optical flow techniques and dynamic programming to estimate the location of a road, and the use of optical flow techniques to improve the quality of an autonomous robot's obstacle classification. |
常見字詞
activity-based ground models approach Autonomous Navigation Autonomous Robot calculated camera chapter classifier output color Computer Vision corresponding current frame DARPA DARPA Grand Challenge defined definition region distance dynamic programming ego-motion environments estimated feature tracking field of view Figure figure shows find first global grid cell horizontal templates IEEE image gradients image plane image registration image segmentation improve Input frame International Conference LAGR learned activity map learned activity models line coverage metric Mahalanobis distance Mobile Robot motion moving platform MRF classifier multi-object tracking occupancy grid off-road OpenCV optical flow field optical flow techniques performance physical bumper pixel coverage metric planner position proposed registration reverse optical flow roadway Robot Navigation robot platform robust search line self-supervised learning shown in Fig Specifically SSD response stereo vision template matching texture thesis Thrun traceback traversable vehicle video frame video sequences video taken