Robotic Object Recognition Using Vision and Touch

1: Introduction.- 1.1 The promise of robotics.- 1.2 Improving robotic system performance.- 1.2.1 Active, exploratory sensing.- 1.2.2 Multiple sensors.- 1.2.3 Complex world models.- 1.2.4 Reasoning about the world.- 1.3 System description.- 1.4 Summary.- 2: Model Database.- 2.1 Object models for recognition.- 2.2 Criteria for a recognition model.- 2.3 A hierarchical model data base.- 2.3.1 Object level.- 2.3.2 Component/feature level.- 2.3.3 Surface level.- 2.3.4 Patch level.- 2.3.5 Relational constraints.- 2.4 Creating the models.- 2.5 Computing surface attributes.- 2.6 Classifying surfaces.- 2.7 Computing hole and cavity attributes.- 2.8 Example model.- 3: 2-D Vision.- 3.1 Introduction.- 3.2 Image acquisition.- 3.3 Thresholding.- 3.4 Edge detection.- 3.5 Segmentation.- 3.5.1 Filling in the gaps.- 3.5.2 Region growing.- 3.5.3 Region analysis.- 3.6 Summary.- 4: 3-D Vision.- 4.1 Introduction.- 4.2 Computational stereo.- 4.3 Image acquisition.- 4.4 Camera modeling.- 4.5 Camera calibration.- 4.6 Feature acquisition.- 4.7 Image matching.- 4.8 Depth determination.- 4.9 Interpolation.- 4.10 Summary.- 5: Tactile Sensing.- 5.1 Introduction.- 5.2 Capabilities of tactile sensors.- 5.3 Summary of tactile sensing.- 5.4 Experimental tactile sensor.- 5.5 Organization of tactile processing.- 5.6 Low level tactile processing.- 5.7 Intermediate level tactile processing.- 5.7.1 Exploring regions.- 5.7.2 Surface tracing.- 5.7.3 Hole/cavity tracing.- 5.8 Summary.- 6: Integrating Vision and Touch.- 6.1 Introduction.- 6.2 Composite bicubic surfaces.- 6.3 Building level 0 surfaces.- 6.3.1 Calculating tangent vectors.- 6.3.2 Calculating twist vectors.- 6.4 Building higher level surfaces.- 6.5 Building hole and cavity descriptions.- 6.6 Summary.- 7: Matching.- 7.1 Introduction.- 7.2 Design of the matcher.- 7.3 Model instantiation.- 7.3.1 Discrimination by size.- 7.3.2 Discrimination by gross shape.- 7.3.3 Feature attribute matching.- 7.3.4 Surface attribute matching.- 7.3.5 Classifying surfaces.- 7.3.6 Relational consistency.- 7.3.7 Ordering matches.- 7.4 Verification.- 7.4.1 Computing model-to-scene transformations.- 7.4.2 Matching feature frames.- 7.4.3 Matching surface frames.- 7.4.4 Verification sensing.- 7.5 Summary.- 8: Experimental Results.- 8.1 Introduction.- 8.2 Experiment 1.- 8.3 Experiment 2.- 8.4 Experiment 3.- 8.5 Experiment 4.- 8.6 Experiment 5.- 8.7 Experiment 6.- 8.8 Experiment 7.- 8.9 Summary.- 9: Conclusion.- 9.1 Introduction.- 9.2 Discovering 3-D structure.- 9.3 The multi-sensor approach.- 9.4 Limitations of the system.- 9.5 Future directions.- References.- Appendix: Bicubic Spline Surfaces.- 1. Introduction.- 2. Parametric curves and surfaces.- 3. Coons' patches.- 3.1 Linearly interpolated patches.- 3.2 Hermite interpolation.- 3.3 Curvature continuous patches.