TY - JOUR T1 - Visual tracking and recognition using appearance-adaptive models in particle filters JF - IEEE Transactions on Image Processing Y1 - 2004 A1 - Zhou,Shaohua Kevin A1 - Chellapa, Rama A1 - Moghaddam, B. KW - adaptive filters KW - adaptive noise variance KW - algorithms KW - appearance-adaptive model KW - Artificial intelligence KW - Cluster Analysis KW - Computer Graphics KW - Computer simulation KW - Feedback KW - Filtering KW - first-order linear predictor KW - hidden feature removal KW - HUMANS KW - Image Enhancement KW - Image Interpretation, Computer-Assisted KW - image recognition KW - Information Storage and Retrieval KW - Kinematics KW - Laboratories KW - Male KW - Models, Biological KW - Models, Statistical KW - MOTION KW - Movement KW - Noise robustness KW - Numerical Analysis, Computer-Assisted KW - occlusion analysis KW - Particle filters KW - Particle tracking KW - Pattern Recognition, Automated KW - Predictive models KW - Reproducibility of results KW - robust statistics KW - Sensitivity and Specificity KW - Signal Processing, Computer-Assisted KW - State estimation KW - statistical analysis KW - Subtraction Technique KW - tracking KW - Training data KW - visual recognition KW - visual tracking AB - We present an approach that incorporates appearance-adaptive models in a particle filter to realize robust visual tracking and recognition algorithms. Tracking needs modeling interframe motion and appearance changes, whereas recognition needs modeling appearance changes between frames and gallery images. In conventional tracking algorithms, the appearance model is either fixed or rapidly changing, and the motion model is simply a random walk with fixed noise variance. Also, the number of particles is typically fixed. All these factors make the visual tracker unstable. To stabilize the tracker, we propose the following modifications: an observation model arising from an adaptive appearance model, an adaptive velocity motion model with adaptive noise variance, and an adaptive number of particles. The adaptive-velocity model is derived using a first-order linear predictor based on the appearance difference between the incoming observation and the previous particle configuration. Occlusion analysis is implemented using robust statistics. Experimental results on tracking visual objects in long outdoor and indoor video sequences demonstrate the effectiveness and robustness of our tracking algorithm. We then perform simultaneous tracking and recognition by embedding them in a particle filter. For recognition purposes, we model the appearance changes between frames and gallery images by constructing the intra- and extrapersonal spaces. Accurate recognition is achieved when confronted by pose and view variations. VL - 13 SN - 1057-7149 CP - 11 M3 - 10.1109/TIP.2004.836152 ER -