Robust visual tracking based on interactive multiple model particle filter by integrating multiple cues

Abstract

Visual tracking can be formulated as a state estimation problem of target representation based on observations in image sequences. To investigate the integration of rough models from multiple cue and to explore computationally efficient algorithms, this paper formulates the problem of multiple cue integration and tracking to combine Interactive Multiple Model (IMM) with particle filter (IMM_PF). Interactive Multiple Model can estimate the multiple cue state of a dynamic system with several behavioral models that switch from one to another using model likelihoods and model transition probabilities. For the problem of visual tracking, the model of IMM is adopted to three target observation models: Corrected Background Weighted Histogram (CBWH), Completed Local Ternary Patterns (CLTP) and Histogram of Oriented Gradients (HOG). The probabilities of these models are corresponding to the weights of multiple cues. IMM_PF then dynamically adjusts the weights of different features. Compared with those state-of-the-art methods in the tracking literature, this algorithm can track the object accurately in conditions of rotation, abrupt shifts, as well as clutter and partial occlusions occurring to the tracking object with good robustness, as demonstrated by experimental results.