Learning Semantic Scene Models by Trajectory Analysis
In this paper, we describe an unsupervised learning framework to segment a scene into semantic regions and to build semantic scene models from long-term observations of moving objects in the scene. First, we introduce two novel similarity measures for comparing trajectories in far-field visual surve...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Language: | en_US |
| Published: |
2006
|
| Online Access: | http://hdl.handle.net/1721.1/31208 |
| _version_ | 1826194718609899520 |
|---|---|
| author | Wang, Xiaogang Tieu, Kinh Grimson, Eric |
| author2 | Eric Grimson |
| author_facet | Eric Grimson Wang, Xiaogang Tieu, Kinh Grimson, Eric |
| author_sort | Wang, Xiaogang |
| collection | MIT |
| description | In this paper, we describe an unsupervised learning framework to segment a scene into semantic regions and to build semantic scene models from long-term observations of moving objects in the scene. First, we introduce two novel similarity measures for comparing trajectories in far-field visual surveillance. The measures simultaneously compare the spatial distribution of trajectories and other attributes, such as velocity and object size, along the trajectories. They also pro-vide a comparison confidence measure which indicates how well the measured im-age-based similarity approximates true physical similarity. We also introduce novel clustering algorithms which use both similarity and comparison confidence. Based on the proposed similarity measures and clustering methods, a framework to learn semantic scene models by trajectory analysis is developed. Trajectories are first clustered into vehicles and pedestrians, and then further grouped based on spatial and velocity distributions. Different trajectory clusters represent different activities. The geometric and statistical models of structures in the scene, such as roads, walk paths, sources and sinks, are automatically learned from the trajectory clusters. Abnormal activities are detected using the semantic scene models. The system is robust to low-level tracking errors. |
| first_indexed | 2024-09-23T10:00:44Z |
| id | mit-1721.1/31208 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T10:00:44Z |
| publishDate | 2006 |
| record_format | dspace |
| spelling | mit-1721.1/312082019-04-11T06:23:38Z Learning Semantic Scene Models by Trajectory Analysis Wang, Xiaogang Tieu, Kinh Grimson, Eric Eric Grimson Vision In this paper, we describe an unsupervised learning framework to segment a scene into semantic regions and to build semantic scene models from long-term observations of moving objects in the scene. First, we introduce two novel similarity measures for comparing trajectories in far-field visual surveillance. The measures simultaneously compare the spatial distribution of trajectories and other attributes, such as velocity and object size, along the trajectories. They also pro-vide a comparison confidence measure which indicates how well the measured im-age-based similarity approximates true physical similarity. We also introduce novel clustering algorithms which use both similarity and comparison confidence. Based on the proposed similarity measures and clustering methods, a framework to learn semantic scene models by trajectory analysis is developed. Trajectories are first clustered into vehicles and pedestrians, and then further grouped based on spatial and velocity distributions. Different trajectory clusters represent different activities. The geometric and statistical models of structures in the scene, such as roads, walk paths, sources and sinks, are automatically learned from the trajectory clusters. Abnormal activities are detected using the semantic scene models. The system is robust to low-level tracking errors. 2006-02-10T22:46:57Z 2006-02-10T22:46:57Z 2006-02-10 MIT-CSAIL-TR-2006-008 http://hdl.handle.net/1721.1/31208 en_US Massachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory 16 p. 18317255 bytes 2106560 bytes application/postscript application/pdf application/postscript application/pdf |
| spellingShingle | Wang, Xiaogang Tieu, Kinh Grimson, Eric Learning Semantic Scene Models by Trajectory Analysis |
| title | Learning Semantic Scene Models by Trajectory Analysis |
| title_full | Learning Semantic Scene Models by Trajectory Analysis |
| title_fullStr | Learning Semantic Scene Models by Trajectory Analysis |
| title_full_unstemmed | Learning Semantic Scene Models by Trajectory Analysis |
| title_short | Learning Semantic Scene Models by Trajectory Analysis |
| title_sort | learning semantic scene models by trajectory analysis |
| url | http://hdl.handle.net/1721.1/31208 |
| work_keys_str_mv | AT wangxiaogang learningsemanticscenemodelsbytrajectoryanalysis AT tieukinh learningsemanticscenemodelsbytrajectoryanalysis AT grimsoneric learningsemanticscenemodelsbytrajectoryanalysis |