Real-Time Tracking via On-line Boosting
Abstract
Very recently tracking was approached using classification techniques such as support vector machines. The object to be tracked is discriminated by a classifier from the background. In a similar spirit we propose a novel on-line AdaBoost feature selection algorithm for tracking. The distinct advantage of our method is its capability of on-line training. This allows to adapt the classifier while tracking the object. Therefore appearance changes of the object (e.g. out of plane rotations, illumination changes) are handled quite naturally. Moreover, depending on the background the algorithm selects the most discriminating features for tracking resulting in stable tracking results. By using fast computable features (e.g. Haar-like wavelets, orientation histograms, local binary patterns) the algorithm runs in real-time. We demonstrate the performance of the algorithm on several (publically available) video sequences.
Principle
The tracking problem is formulated as binary classification problem. Since the object itself and/or the surrounding background can change over time we describe the object by an on-line discriminative classifier. The representation is based on the well known Viola/Jones object detector but now with the capability to change it by new samples. This is done by using a on-line boosting approach (see On-line Boosting and Vision for dome more details).
- The presentation given at BMVC'06 is availible here as PDF-document.
Videos
- Tracking the Invisible
- My girlfriend and I...
- Some videos from David Ross' project webpage :
Demonstation Program
A binary with a simple GUI demonstrating the proposed tracker (without scale adaption and no further postprocessing, like trajectory smoothing)
- coming coon...
