Forschungsbericht 09-18: Meysel F. Map-based multi-object dynamic scene: description and atypical event detection / Deutsches Zentrum für Luft- und Raumfahrt Institut für Flugfühurung, Braunschweig. - Köln: DLR, Bibliotheks- und Informationswesen, 2009. - xx, 215 p.: ill., graph. - Bibliogr.: p.207-215. - ISSN 1434-8454  Место хранения:   061 | Институт оптики атмосферы CO РАН | Томск | Библиотека

Contents

1. Introduction ................................................. 1 1.1. Motivation .............................................. 3 1.2. Background .............................................. 4 1.3. Aims of the thesis ...................................... 6 1.4. Structure of the thesis ................................. 7 2. Related Work ................................................ 11 2.1. Automatic situation understanding ...................... 11 2.2. Recurring key concepts ................................. 20 2.3. Modeling the scene ..................................... 33 2.4. Summary of related work ................................ 40 3. Method ...................................................... 43 3.1. Model: The scene controls the behavior ................. 43 3.2. Data structures storing scene dynamics ................. 54 3.2.1. Level 0 maps .................................... 54 3.2.2. Level I maps .................................... 57 3.2.3. Level II maps ................................... 60 3.3. Detection of scene changes (MBCD) and atypical behavior (MBTA) ........................................ 65 3.4. Differences to other approaches ........................ 68 3.5. Summary of Method ...................................... 71 4. Implementation and Experiment ............................... 73 4.1. Campaign data used for implementation .................. 73 4.2. Data acquisition and preprocessing ..................... 76 4.3. Scene dynamics acquisition ............................. 80 4.4. Map based scene change detection ....................... 85 4.5. Atypical event detection ............................... 87 4.6. Human decisions in atypical event detection ............ 95 4.7. Implementation summary ................................. 98 5. Results and Comparison ..................................... 101 5.1. Scene change detection ................................ 102 5.1.1. MESS ........................................... 103 5.1.2. STUT ........................................... 104 5.1.3. PRAG ........................................... 106 5.2. Atypical event detection .............................. 107 5.2.1. MESS ........................................... 108 5.2.2. STUT ........................................... 115 5.2.3. PRAG ........................................... 117 5.3. Characteristics of datasets ........................... 120 5.4. Interpretation and comparison of results .............. 123 5.4.1. Interpretation of human results ................ 123 5.4.2. Comparison of human and MBCD results .......... 126 5.4.3. Comparison of human and MBTA results ........... 128 5.5. Summary of results .................................... 131 6. Discussion ................................................. 133 6.1. Map-based change detection MBCD ....................... 133 6.2. Map-based trajectory analysis MBTA .................... 135 6.3. Strengths and limitations of MBCD and MBTA ............ 136 6.4. Revisiting assertions of the thesis ................... 142 6.5. Outlook and future work ............................... 143 6.6. Summary ............................................... 146 A Moving Object Extraction .................................. 149 В Photogrammetric Model ..................................... 165 С Derivation of Level-ll maps ............................... 171 D Activity, speed and heading maps .......................... 177 E Colortable for headings ................................... 187 F Map based scene change result plots ....................... 189 G Instructions for human probands ........................... 195 H Results from human atypical event tagging ................. 199