机器人技术应用论文.pptx

3-D Vision and Recognition ;CONTENTS;The develop of vision;The develop of vision;The develop of vision;Application of computer Vision;Application of computer Vision;Application of computer Vision;Application of computer Vision;Application of computer Vision;3D sensor;Laser radar;TOF CAMERA;STRUCTURE CAMERA;3-D Vision and Visual SLAM;3-D Vision and Visual SLAM;Let us start by introducing the projection of the world to an image plane.;According to the similarity of triangles, the relationship between camera coordinates and image coordinates as follow;3-D Vision and Visual SLAM;Pose Estimation Solution;We assume that a camera is calibrated and that measurements of N points are given in world coordinates Xj=1,... ,N and calibrated image coordinates xj=1,... ,N.;Pose Estimation Solution;Triangulation;Moving Stereo;Structure from Motion : Epipolar Geometry;To avoid the scale ambiguity we assume that t is a unit vector. We proceed by summarizing the unknowns into one matrix;3-D Vision and Visual SLAM;3-D Vision and Visual SLAM;Recognition;Approaches to Recognition;Appearance-Based Methods;Appearance-Based Methods : Distinctive Features;Appearance-Based Methods : Distinctive Regions;Appearance-Based Methods : Descriptors for Distinctive Points and Regions;Matching : Vocabulary-Based Methods;Recognition Using Vocabulary Trees;Place Recognition and Terrain Classi?cation;SLAM;Vision-based Simultaneous Localization and Mapping in Changing Outdoor Environments;CONTENTS;INTRODUCTION;INTRODUCTION;APPROACH;APPROACH;APPROACH : RatSLAM;APPROACH : Patch-based Visual Odometry;APPROACH : Filtered Intensity Pro?le Visual Odometry;APPROACH : Filtered Intensity Pro?le Visual Odometry;APPROACH : Patch-normalized Whole Image Place Recognition;APPROACH : Patch-normalized Whole Image Place Recognition;APPROACH : Patch Veri?cation;APPROACH : Saliency Mask;APPROACH : Saliency Mask;RESULTS;RESULTS