Inhalt

•  
   Abstract
•  
   Contents
•  
   1 Introduction
•  
   1.1 Motivation for Road Terrain Detection
•  
   1.2 Restriction of State-of-the-Art
•  
   1.3 Basic Concept of the Proposed Approach
•  
   1.4 Contribution to State-of-the-Art
•  
   1.5 Outline
•  
   2 Related Work
•  
   2.1 Approaches for Road Terrain Detection
•  
   2.1.1 Delimiter-based Road Terrain Detection
•  
   2.1.2 Surface-based Road Terrain Detection
•  
   2.1.3 Inclusion of Top-down Scene Context
•  
   2.2 Approaches employing Spatial Features
•  
   2.3 Summary
•  
   3 Road Terrain Detection System
•  
   3.1 System Overview
•  
   3.1.1 Representation of Local Visual Appearance
•  
   3.1.2 Inclusion of the Spatial Layout of Visual Appearance
•  
   3.2 Summary
•  
  4 Road Terrain Detection using Local Visual Features
•  
   4.1 System for Local Visual Appearance Classification
•  
  4.2 Extraction of Local Visual Appearance using Patch-based Features
•  
   4.2.1 Color Features
•  
   4.2.2 Walsh Hadamard Texture Features
•  
   4.2.3 Learning Road Appearance Descriptors using Slow Feature Analysis
•  
   4.3 Appearance Classification using Boosting
•  
   4.3.1 Training of GentleBoost
•  
   4.3.2 Classifier Processing and Back Projection
•  
  4.4 Dedicated Generation of Training Data
•  
   4.4.1 Generation of Training Data for Road Area
•  
   4.4.2 Generation of Training Data for Road Boundary
•  
   4.5 Experiments and Results of Local Visual Appearance based Classification
•  
   4.5.1 Perspective and Metric Evaluation Measures
•  
   4.5.2 Influence of Local and Global Image Normalization
•  
   4.5.3 Influence of Feature Order and Different Feature Combinations
•  
   4.5.4 Evaluation and Results of Road Area Detection
•  
   4.5.5 Results of Road Boundary Detection
•  
   4.6 Summary
•  
  5 Incorporating the Spatial Layout of Local Visual Appearance
•  
   5.1 Basic Concept for Visuospatial Classification
•  
   5.2 Base Classification for Capturing Local Visual Appearance
•  
   5.3 Spatial Layout Computation
•  
   5.3.1 Method for Spatial Layout Computation
•  
   5.3.2 Spatial Ray (SPRAY) Feature Generation
•  
   5.3.3 SPRAY Algorithm
•  
   5.3.4 Combining SPRAY Features to a Visuospatial Representation
•  
   5.3.5 Discrimination of Ego-Lane from other Lanes
•  
   5.4 Road Terrain Classification
•  
   5.5 Experiments on Road Area Detection
•  
   5.5.1 Assessing Single Cue Road Area Classifier Performance
•  
   5.5.2 Influence of Road Area Classifier Cue Combinations
•  
   5.5.3 Qualitative Results
•  
   5.6 Experiments on Ego-Lane Detection
•  
   5.6.1 Assessing Single Cue Ego-Lane Classifier Performance
•  
   5.6.2 Influence of Ego-Lane Classifier Cue Combinations
•  
   5.6.3 Qualitative Results
•  
   5.7 Summary
•  
   6 Visual Ego-Vehicle Lane Assignment using Spatial Ray Features
•  
   6.1 Related Work for Ego-vehicle Localization
•  
   6.2 System Description
•  
   6.3 Lane Index Classification
•  
   6.4 Evaluation
•  
   6.4.1 Left and Right Ego-Lane Index Classification (Highway 1)
•  
   6.4.2 Generalization Experiments (Highway 2)
•  
   6.4.3 Discussion
•  
   6.5 Summary
•  
  7 State-of-the-Art Comparison and Future Embedding of the Concept
•  
   7.1 Road Area Detection Performance Comparison with State-of-the-Art
•  
   7.2 Enhanced Robustness Through Adaptation
•  
   7.3 Summary
•  
   8 Summary and Conclusion
•  
   References
•  
   A Inverse Perspective Mapping
•  
  B Inner-city Evaluation Datasets
•  
   B.1 Benchmark Dataset
•  
   B.2 Sequential Inner-city Dataset
•  
   C Baseline Performance Evaluation
•  
   D Evaluation of Road Boundary Detection