|
||
|
|
|
|
 |
||||||
Home |
Deformable Contour Tracking
using Hidden Markov Models Members: Kyle Brocklehurst, Chen-Ping Yu, Sitapa Rujikietgumjorn Abstract Object tracking is useful in many applications, and it is an active research area of computer vision. It is commonly used in human computer interaction and visual surveillance systems. Conventional methods such as background subtraction and mode seeking have been widely used, while many new approaches involving active shape models and graphical models were proposed over the past few years. However, these approaches usually use a pre-defined template for matching, which allows very little deformation. In this project, we focused on object tracking using contours. Specifically, we attempt to track objects that may deform substantially. Where this can be useful is for image segmentation and general object tracking. We formulate the contour as a Hidden Markov Model. In each new frame, we allow the states (locations along the boundary) to adapt to new observations of edge detection and NCC score against a patch centered at their location along the previous frame, while constraining smoothness of neighboring transitions as well. To do this, the forward-backward algorithm is used to find the global optimum set of contour landmark points in the new frame. [PDF]  [AVI]Tracking an object with translation and slight deformation such as turning your fist is fairly accurate and can recover even after temporarily losing the contour.  [AVI]Our method is capable of tracking complex deformations and maintains accuracy by adjusting the number of landmark points as the surface area increases.  [AVI]This shows a failure of our algorithm. The deformation here is too extreme and happening too quickly. This represents an area where either the cost weighting parameters or the similarity and edge measures could be improved. |
|||||
 |
||
|
