Details
| Original language | English |
|---|---|
| Pages (from-to) | 145-150 |
| Number of pages | 6 |
| Journal | International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives |
| Volume | 40 |
| Issue number | 3 |
| Publication status | Published - 14 Aug 2014 |
| Event | ISPRS Technical Commission III Symposium 2014 - Zurich, Switzerland Duration: 5 Sept 2014 → 7 Sept 2014 |
Abstract
Many tracking systems rely on independent single frame detections that are handled as observations in a recursive estimation framework. If these observations are imprecise the generated trajectory is prone to be updated towards a wrong position. In contrary to existing methods our novel approach suggests a Dynamic Bayes Network in which the state vector of a recursive Bayes filter, as well as the location of the tracked object in the image are modelled as unknowns. These unknowns are estimated in a probabilistic framework taking into account a dynamic model, prior scene information, and a state-of-the-art pedestrian detector and classifier. The classifier is based on the Random Forests-algorithm and is capable of being trained incrementally so that new training samples can be incorporated at runtime. This allows the classifier to adapt to the changing appearance of a target and to unlearn outdated features. The approach is evaluated on a publicly available dataset captured in a challenging outdoor scenario. Using the adaptive classifier, our system is able to keep track of pedestrians over long distances while at the same time supporting the localisation of the people. The results show that the derived trajectories achieve a geometric accuracy superior to the one achieved by modelling the image positions as observations.
Keywords
- Classification, On-line, Reasoning, Tracking, Video
ASJC Scopus subject areas
- Computer Science(all)
- Information Systems
- Social Sciences(all)
- Geography, Planning and Development
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives, Vol. 40, No. 3, 14.08.2014, p. 145-150.
Research output: Contribution to journal › Conference article › Research › peer review
}
TY - JOUR
T1 - A dynamic Bayes Network for visual pedestrian tracking
AU - Klinger, T.
AU - Rottensteiner, F.
AU - Heipke, C.
PY - 2014/8/14
Y1 - 2014/8/14
N2 - Many tracking systems rely on independent single frame detections that are handled as observations in a recursive estimation framework. If these observations are imprecise the generated trajectory is prone to be updated towards a wrong position. In contrary to existing methods our novel approach suggests a Dynamic Bayes Network in which the state vector of a recursive Bayes filter, as well as the location of the tracked object in the image are modelled as unknowns. These unknowns are estimated in a probabilistic framework taking into account a dynamic model, prior scene information, and a state-of-the-art pedestrian detector and classifier. The classifier is based on the Random Forests-algorithm and is capable of being trained incrementally so that new training samples can be incorporated at runtime. This allows the classifier to adapt to the changing appearance of a target and to unlearn outdated features. The approach is evaluated on a publicly available dataset captured in a challenging outdoor scenario. Using the adaptive classifier, our system is able to keep track of pedestrians over long distances while at the same time supporting the localisation of the people. The results show that the derived trajectories achieve a geometric accuracy superior to the one achieved by modelling the image positions as observations.
AB - Many tracking systems rely on independent single frame detections that are handled as observations in a recursive estimation framework. If these observations are imprecise the generated trajectory is prone to be updated towards a wrong position. In contrary to existing methods our novel approach suggests a Dynamic Bayes Network in which the state vector of a recursive Bayes filter, as well as the location of the tracked object in the image are modelled as unknowns. These unknowns are estimated in a probabilistic framework taking into account a dynamic model, prior scene information, and a state-of-the-art pedestrian detector and classifier. The classifier is based on the Random Forests-algorithm and is capable of being trained incrementally so that new training samples can be incorporated at runtime. This allows the classifier to adapt to the changing appearance of a target and to unlearn outdated features. The approach is evaluated on a publicly available dataset captured in a challenging outdoor scenario. Using the adaptive classifier, our system is able to keep track of pedestrians over long distances while at the same time supporting the localisation of the people. The results show that the derived trajectories achieve a geometric accuracy superior to the one achieved by modelling the image positions as observations.
KW - Classification
KW - On-line
KW - Reasoning
KW - Tracking
KW - Video
UR - http://www.scopus.com/inward/record.url?scp=84924263554&partnerID=8YFLogxK
U2 - 10.5194/isprsarchives-XL-3-145-2014
DO - 10.5194/isprsarchives-XL-3-145-2014
M3 - Conference article
AN - SCOPUS:84924263554
VL - 40
SP - 145
EP - 150
JO - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives
JF - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives
SN - 1682-1750
IS - 3
T2 - ISPRS Technical Commission III Symposium 2014
Y2 - 5 September 2014 through 7 September 2014
ER -