Mobile robots require the ability to make decisions such as "go through the hedges" or "go around the brick wall." Mobile Robot Navigation with Intelligent Infrared Image Interpretation describes in detail an alternative to GPS navigation: a physics-based adaptive Bayesian pattern classification model that uses a passive thermal infrared imaging system to automatically characterize non-heat generating objects in unstructured outdoor environments for mobile robots. The resulting classification model complements an autonomous robot’s situational awareness by providing the ability to classify smaller structures commonly found in the immediate operational environment.
Author(s): William L. II Fehlman, Mark K. Hinders
Edition: 1st Edition.
Year: 2009
Language: English
Pages: 274
Contents......Page 28
1.1 Purpose of Book......Page 31
1.2 Non-Heat Generating Objects......Page 36
1.3 Autonomous Robotic Systems......Page 38
1.3.1 Detect the Object......Page 41
1.3.3 Classify the Object......Page 43
1.4 Infrared Thermography......Page 45
1.4.1 Active vs. Passive Thermography......Page 46
1.4.2 Advantages & Disadvantages of Thermal Infrared Imaging......Page 47
1.4.3 Multi-Mode Heat Transfer Model......Page 48
1.5 Overview of the Book......Page 50
References......Page 52
2.2.1 Hardware......Page 55
2.2.2 Signal Preprocessing......Page 58
2.3 Data Collection......Page 71
References......Page 75
3.1 Introduction......Page 77
3.2 "Ugly Duckling" Features......Page 78
3.3 Thermal Image Representation......Page 81
3.4.2 Ambient Temperature Rate of Change......Page 84
3.5.1 Emissivity Variation by Material Type......Page 85
3.5.2 Emissivity Variation by Viewing Angle......Page 88
3.5.4 Emissivity Variation by Shape and Surface Temperature......Page 89
3.5.5 Other Directional Variation Enhancers......Page 90
3.5.6 Emissivity-based Features......Page 93
3.6 Macro Features......Page 99
3.6.1 First-order Statistical Features......Page 100
3.6.2 Second-order Statistical Features......Page 104
3.7 Thermal Feature Application......Page 114
3.8 Curvature Algorithm......Page 118
References......Page 121
4.1 Introduction......Page 125
4.2 "No Free Lunch" Classifiers......Page 126
4.3 Preliminary Feature Analysis......Page 128
4.4.1 Bayesian Classifier......Page 135
4.4.2 K-Nearest-Neighbor (KNN) Classifier......Page 140
4.4.3 Parzen Classifier......Page 141
4.4.4 General Remarks......Page 143
4.5 Model Performance and Feature Selection......Page 145
4.5.1 Feature Selection Method......Page 146
4.5.2 Performance Criterion......Page 150
4.5.3 Error Estimation Method......Page 152
4.5.5 Extended Object Performance and Feature Selection......Page 153
4.5.6 Compact Object Performance and Feature Selection......Page 163
4.6 Sensitivity Analysis......Page 173
4.6.1 Viewing Angle Variations......Page 174
4.6.2 Window Size Variations......Page 178
4.6.3 Rotational Variations......Page 185
4.7 Summary......Page 186
References......Page 188
5.1 Introduction......Page 191
5.2 Distance Metrics for Hyperconoidal Clusters......Page 192
5.3 Adaptive Bayesian Classifier Design......Page 223
5.4.1 Blind Data Performance......Page 227
5.4.2 Analysis of Misclassifications......Page 240
5.5 Adaptive Bayesian Classification Model Design......Page 253
5.6 Adaptive Bayesian Classification Model Application......Page 257
5.6.1 Performance on Blind Data (with Classes = Training Set)......Page 258
5.6.2 Performance on Blind Data (with Classes ≠ Training Set)......Page 277
5.7 Summary......Page 281
References......Page 284
6.1 Introduction......Page 285
6.2 Contributions......Page 286
6.3 Limitation of a Thermal Infrared Imaging System......Page 287
6.4.1 Augmentation of Robotic Thermal Imaging System......Page 289
6.4.2 Fuzzy Logic Classifier......Page 290
6.4.3 Bayesian Multi-Sensor Data Fusion......Page 295
6.4.4 Prior Knowledge Based on Satellite Imagery......Page 297
6.5 Concluding Remarks......Page 298
References......Page 299
E......Page 301
N......Page 302
T......Page 303
Z......Page 304
