Object Tracking

Project member: Taoran Lu

Advisor: Prof. Dapeng Oliver Wu

|| Introduction || Results || Downloads || Useful Links || Future Work || Back to Taoran's Website ||

Object tracking, by definition, is to track an object (or multiple objects) over a sequence of images. Object tracking, in general, is a challenging problem. Difficulties in tracking objects can arise due to abrupt object motion, changing appearance patterns of both the object and the scene, non-rigid object structures, object-to-object and object-to-scene occlusions, and camera motion. Tracking is usually performed in the context of higher-level applications that require the location and/or shape of the object in every frame.

Generally there are three types of tracking algorithm:

In our research, we studied feature-based point tracking and meanshift kernel tracking.

An introduction to meanshift tracking algorithm

An introduction to feature-based tracking algorithm

We implemented these two algorithms to a MATLAB graphical user interface. The program loads a set of qcif video sequences as input and displays the tracking result.  User can set arbitrary frame as beginning or ending frame by previewing the frames before tracking. Two output file types are supported: avi video clip and animated gif image. 

Demo1: Meanshift tracking: (Please wait while media is loading and please make sure your Explorer does not block activeX contents)

If you cannot see the demos, you can also download them here:

Meanshift tracking demo and Feature tracking demo

To achieve best visual quality, please set your screen resolution higher than 1280*1024.

Demo2: Feature-based tracking:

Matlab source code: TrackingGui.zip

Test video sequences can be find at: http://media.xiph.org/video/derf/

A. Yilmaz, O. Javed and M. Shah, "Object Tracking: A Survey," ACM Journal of Computing Surveys, Vol. 38, No. 4, 2006

Object detection and tracking

Based on the tracking test bench, we develop an objective tracking assessment algorithm. The assessment algorithm is focused on two major criteria:

(1) If the tracking algorithm track the target object. (T/F problem)

There are three conditions:

Hit: Track the right object.

Miss: Track the background.

 Wrong: Track the other object(s) if there are multiple objects.

(2) How good the tracking algorithm works. (Accuracy problem)

The accuracy is evaluated by the distance from the tracked centroid to the ground truth.

to be continued..

Last modified 07/14/2010 16:30:49        Copyright @ Taoran Lu