A Structural Analysis of Complex Aerial Photographs

1. Introduction.- 1.1 Preliminary Remarks.- 1.2 Statistical Classification Methods.- 1.3 Target Detection.- 1.4 Image Understanding Applied to Aerial Photographs.- 1.4.1. Knowledge Sources in the Analysis of Aerial Photographs.- 1.4.2. Problems in Automatic Photointerpretation.- 1.5 Structural Analysis of Complex Aerial Photographs.- 2. Overview of the System.- 2.1 Specifications of the Aerial Photographs Under Analysis.- 2.2 Processing Sequence in the System.- 2.3 Production System as the Software Architecture.- 2.3.1. Methodology in a Production System.- 2.3.2. Production System in the Analysis of Aerial Photographs.- 2.4 Focusing Mechanism of the System.- 3. Some Basic Techniques in Picture Processing and Feature Extraction.- 3.1 Edge-Preserving Smoothing.- 3.1.1. Blurring Effect of Smoothing.- 3.1.2. Edge-Preserving Smoothing.- 3.1.3. Actual Implementation for Discrete Picture Data.- 3.1.4. Sharpening of Blurred Edges.- 3.1.5. Convergence.- 3.1.6. Conclusion.- 3.2 A Measure of Elongatedness of a Region.- 3.2.1. Elongatedness of a Region.- 3.2.2. Longest Path Detection Algorithm.- 3.2.3. Calculation of the Elongatedness.- 3.3 A Structural Description of Regularly Arranged Patterns.- 3.3.1. Introduction.- 3.3.2. Describing Spatial Relationships Using Relative Vectors Among Elements.- 3.3.3. Extraction of Regularity Vectors.- 3.3.4. Estimating Locations of Missing Elements.- 3.3.5. Describing Spatial Arrangements of Elements.- 3.3.5.1. One-Dimensional Repetitive Pattern.- 3.3.5.2. Two-Dimensional Lattice.- 3.3.5.3. Hierarchical Arrangement.- 3.3.6. Discussion.- 3.3.7. Conclusion.- 4. Structuring of Picture Data.- 4.1. Low-Level Processing in Image Understanding.- 4.2. Edge-Preserving Smoothing.- 4.3. Segmentation.- 4.3.1. Merging of Pixels and Labeling of Elementary Regions.- 4.3.2. Threshold Determination.- 4.4. Calculation of Basic Properties of Elementary Regions.- 5. Extraction of Characteristic Regions.- 5.1. Large Homogeneous Regions.- 5.2. Elongated Regions.- 5.3. Shadow Regions and Shadow-Making Regions.- 5.4. Vegetation Regions.- 5.5. Water Regions.- 5.6. High-Contrast Texture Areas.- 6. Object Recognition.- 6.1. Crop Field.- 6.2. Forest and Grassland.- 6.3. Road and River.- 6.4. Car.- 6.5. House and Building.- 6.5.1. House1.- 6.5.2. House2.- 6.5.3. House3.- 6.5.4. House4.- 6.5.5. Building.- 7. Control Structure of the System.- 7.1. Structure of the Blackboard.- 7.1.1. The Global Parameter Table.- 7.1.2. The Property Table.- 7.1.3. Label Picture.- 7.2. Conflict Resolution.- 7.3. Correction of Segmentation Errors 162 "Bottle-neck" detection algorithm.- 8. Performance Evaluation.- 8.1. Some Examples of the Analysis.- 8.2. Processing Time Evaluation.- 9. Conclusion.- 9.1. Summary.- 9.2. Areas for Future Work.- Overview of the Authors' Laboratory.- Computer Facilities.