Under the framework of sequential Monte Carlo filtering algorithm, we try to get insights on some key issues in visual tracking with application scenarios on face and human tracking. Some important technologies and solutions are studied which are necessary for robust and practical tracking systems, especially concentrate on how to model the object appearance variation, how to model and estimate the object complex dynamics, how to combine low level and high level motion estimation methods to enhance tracker robust and efficiency.

• Adaptive Discriminative Object Appearance Model for Object Tracking

We have proposed an algorithm for online modeling and adapting discriminative object appearance model. The algorithm considers both object appearance and its relevant background when constructing object model. The constructed model encodes the difference between the object and background dynamically. Therefore, during the tracking process, the discriminability of the updated model is guaranteed basically. Compared with one of the state-of-the-art tracking algorithms, Mean Shift, experimental results show that our algorithm performs better on publicly available test sequences. See our CVPR2005 paper or contact Dr.Jianyu Wang.

• Multiple Sequential Monte Carlo Filter for Complex Motion

We have proposed a framework to model and estimate complex motion by incorporating multiple motion models. Based on the proposed framework and aim to solving two kinds of complex motions, two new variations of sequential Monte Carlo filter are proposed, termed as multi-model switching sequential Monte Carlo filter and multi-model cooperation sequential Monte Carlo filter respectively, by combining the proposed framework with the standard sequential Monte Carlo filter. Experimental results show that the proposed algorithms perform better than standard sequential Monte Carlo filter and simultaneously lower the computational burden. See our PCM2004 paper  or contact Dr.Jianyu Wang.