Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 43664 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Jamzad, Mansour
- Abstract:
- To overcome challenges such as object appearance changes and environment illumination variations in tracking methods, online algorithms are suggested to be used instead of offline ones. Online algorithms update the model by the information acquired in the last processed frame. The main challenge of using online algorithms is the accumulation of small errors after several steps of updating of the model (drift) which disturbs the model and causes tracking failure. Using the object information in the first frame in each update can be considered as a solution. The proposed online semi-supervised boosting algorithms can overcome the drift problem at the expense of decreasing their capabilities in handling object appearance changes challenges.In this thesis we propose two online methods for object tracking. In the first method, the object is modelled by a discrete distribution. In each frame,color distribution of target candidates is obtained and the candidate having the lowest distance to the object distribution is considered as the object. Then, the model is updated using the discrete distribution clustering algorithm. In this method, particle filter is used for searching the state space. Since the model is updated in each frame, the algorithm is robust with respect to appearance changes and illumination variations. The proposed algorithm can overcome partial and full occlusion as well because it uses object distribution in the first frame in each updating of the model.The second proposed method uses an online semi-supervised boosting algorithm for object tracking. Up to now,the proposed semi-supervised tracking methods only use the object information in the first frame for labeling the unlabeled samples. In this thesis we present a new online semi supervised boosting algorithm which uses the object informationin all of the frames in order to update the model instead of using just the first frame. The method is online and one prototype of all the data which has been seen up to now is maintained. This prototype is updated using the discrete distribution clustering algorithm when new data is received. By using this type of updating the algorithm becomes robust with respect to challenges such as partial and full occlusion, object appearance changes, illumination variations, etc.Experimental results show the robustness of our proposed methods on challenging videos in comparison with other tracking algorithms.
- Keywords:
- Visual Tracking ; Particle Filter ; Online Boosting ; Semisupervised Boosting ; Discrete Distribution Clustering
Digital Object List-
محتواي پايان نامه - view
Bookmark