Скачать с ютуб Lecture 7: Visual Navigation for Flying Robots (Dr. Jürgen Sturm) в хорошем качестве

Lecture 7: Visual Navigation for Flying Robots (Dr. Jürgen Sturm)

Topics covered: RANSAC algorithm Iterative closest point algorithm The SLAM problem Graph SLAM Non-linear map optimization Course website: http://vision.in.tum.de/teaching/ss20... Speaker: Dr. Jürgen Sturm http://vision.in.tum.de/members/sturmju Slide titles: 00:00:10 Agenda for Today 00:01:02 Remember: 8-Point Algorithm 00:01:39 How to deal with outliers? 00:02:12 Robust Estimation 00:03:11 Random Sample Concensus (RANSAC) 00:04:11 Example 00:05:37 How many iterations do we need? 00:07:11 Summary on RANSAC 00:07:43 Laser-based Motion Estimation 00:09:06 Laser Triangulation 00:10:32 Example: Neato XV-11 00:11:19 How does the data look like? 00:12:11 Laser Scanner 00:14:28 Laser-based Motion Estimation 00:15:04 Exhaustive Search 00:15:42 Example: Exhaustive Search 00:17:52 Generalization to 3D 00:19:36 Iterative Closest Point (ICP) 00:20:52 Known Correspondences 00:22:38 Unknown Correspondences 00:23:03 ICP Algorithm 00:23:37 Example: ICP 00:24:10 ICP Variants 00:24:40 Performance Criteria 00:25:20 Selecting Source Points 00:26:20 Spatially Uniform Sampling 00:27:18 Feature-based Sampling 00:28:11 Closest Point Matching 00:29:14 Speeding Up Correspondence Search 00:29:47 Projection-based Matching 00:31:12 Error Metrics 00:33:22 Minimization 00:33:47 Example: Real-Time ICP on Range Images 00:34:59 ICP: Summary 00:35:46 The SLAM Problem 00:37:42 SLAM Applications 00:38:26 SLAM with Ceiling Camera (Samsung Hauzen RE70V) 00:40:00 Localization, Path Planning, Coverage (Neato XV11) 00:40:58 SfM vs. SLAM 00:42:17 Remember: 3D Transformations 00:43:35 Remember: 3D Rotation as Axis/Angle 00:44:41 Remember: 3D Twists 00:45:21 Notation 00:45:46 Incremental Motion Estimation 00:47:16 Loop Closures 00:48:32 Graph SLAM 00:49:08 Example: Graph SLAM on Intel Dataset 00:50:21 Graph SLAM Architecture 00:51:25 Problem Definition 00:52:14 Map Error 00:53:22 Error Function 00:55:40 Non-Linear Optimization Techniques 00:56:14 Gauss-Newton Method 00:56:56 Linearization and Derivation 00:57:34 Linearizing the Error Function 00:58:44 Derivatives of the Error Terms 00:59:42 Linearizing the Error Function 01:00:48 (Linear) Least Squares Minimization 01:01:39 Gauss-Newton Method 01:02:21 Structure of the Minimization Problem 01:03:23 Illustration of the Structure 01:05:20 Sparsity of the Hessian 01:06:21 Example in 1D 01:08:43 What went wrong? 01:09:36 Fixing one node 01:10:06 Levenberg-Marquardt Algorithm 01:17:49 Non-Linear Minimization 01:20:22 Example: Google Street View (Ceres Solver) 01:21:23 Example: RGB-D SLAM 01:25:21 Lessons Learned Today