Tracker Module
(Fixed grammar errors) |
|||
| Line 2: | Line 2: | ||
[[Image:TrackerFramework.png|thumb|center|629px| Figure.1 Framework of the Tracker]] | [[Image:TrackerFramework.png|thumb|center|629px| Figure.1 Framework of the Tracker]] | ||
The tracking system is automatically initialized by the detector in the first two frames. | The tracking system is automatically initialized by the detector in the first two frames. | ||
| − | For separate worms, the single-worm tracker is used. | + | For separate worms, the single-worm tracker is used. When two or more worms come into contact, the multi-worm tracker is initiated. |
| − | Both trackers are based on | + | Both trackers are based on a particle filter. Figure.2 shows 100 particles (predictions of the current state: axial progression and radial displacement) at one frame, and Figure.3 shows the two observation models used for computing the likelihood of particles. The region-based observation model computes likelihoods based on inward and outward intensity information, while the edge-based observation model uses the distances between the model points and detected edge points. |
| − | [[Image:Particles.png|thumb|center|500px| Figure.2 100 particles of one worm at one frame(the head part | + | [[Image:Particles.png|thumb|center|500px| Figure.2 100 particles of one worm at one frame (the head part has a larger amplitude of displacement)]] |
[[Image:ObservationModel.png|thumb|center|550px| Figure.3 Observation models ]] | [[Image:ObservationModel.png|thumb|center|550px| Figure.3 Observation models ]] | ||
| − | After tracking, the center line of worm will be further adjusted, as shown in Figure.4. | + | After tracking, the center line of the worm will be further adjusted, as shown in Figure.4. |
[[Image:CenterLineAdjust.png|thumb|center|550px| Figure.4 Center Line Adjustment based on 1D Filtering]] | [[Image:CenterLineAdjust.png|thumb|center|550px| Figure.4 Center Line Adjustment based on 1D Filtering]] | ||
| − | We made several modifications for the multi-worm trakcer. For more details about its | + | We made several modifications for the multi-worm trakcer. For more details about its differences from the single-worm tracker, please refer to: |
:*[http://www.rpi.edu/~wangy15/Multi-Worm%20Tracking.pdf'''Report for the Tracking System'''] | :*[http://www.rpi.edu/~wangy15/Multi-Worm%20Tracking.pdf'''Report for the Tracking System'''] | ||
Revision as of 14:43, 6 July 2009
The framework for the Multi-Worm Tracker is shown in Figure.1.
The tracking system is automatically initialized by the detector in the first two frames. For separate worms, the single-worm tracker is used. When two or more worms come into contact, the multi-worm tracker is initiated. Both trackers are based on a particle filter. Figure.2 shows 100 particles (predictions of the current state: axial progression and radial displacement) at one frame, and Figure.3 shows the two observation models used for computing the likelihood of particles. The region-based observation model computes likelihoods based on inward and outward intensity information, while the edge-based observation model uses the distances between the model points and detected edge points.
After tracking, the center line of the worm will be further adjusted, as shown in Figure.4.
We made several modifications for the multi-worm trakcer. For more details about its differences from the single-worm tracker, please refer to:
