Research On Image Sparse Representation Model And Its Applications In Visual Tracking

Visual tracking is a well studied issue in computer vision and plays a crucial role in many practical applications such as video surveillance, human motion understanding, and interactive video processing, and so on. Although existing trackers have made some success under various scenarios, objects tracking is still challenging because the appearance of an object can be changed drastically while it undergoes significant pose change, illumination variation and/or partial occlusion. Starting from the sparse representation of images, this dissertation conducts an in-depth study on visual tracking. The main contents are summarized as follows:●An online algorithm is proposed by combining Multiple Instance Learning (MIL) and local sparse representation for tracking an object in a video system. The key idea in our method is to model the appearance of an object by local sparse codes which can be formed as training data for the MIL framework. First, local image patches of a target object are represented as sparse codes with an over-complete dictionary, where the adaptive representation can be helpful to overcoming partial occlusion in object tracking. Then MIL learns the sparse codes by a classifier to discriminate the target from the background. Finally, results from the trained classifier are input into a particle filter framework to sequentially estimate the target state over time in visual tracking. In addition, to decrease the visual drift due to the accumulative errors when updating the dictionary and classifier, a two-step object tracking method combining a static MIL classifier with dynamical MIL classifier is proposed.●A multi-scale patch based appearance model is proposed by means of sparse representation and an efficient scheme involving the collaboration between multi-scale patches is constructed by sparse coefficients. The model exploits both partial and spatial information of the target based on multi-scale patches. Finally, a similarity score of one candidate target is input into a particle filter framework to estimate the target state sequentially over time in visual tracking. Additionally, to decrease the visual drift caused by frequently updating model, we present a novel two-step object tracking method which exploits both the ground truth information of the target labeled in the first frame and the target obtained online with the multi-scale patch information.●A collaborative object tracking model with local sparse representation is proposed. The key idea of the method is to develop a local sparse representation-based discriminative model (SRDM) and a local sparse representation-based generative model (SRGM). In the SRDM module, the appearance of a target is modeled by local sparse codes that can be formed as training data for a linear classifier to discriminate the target from the background. In the SRGM module, the appearance of the target is represented by sparse coding histogram and a sparse coding-based similarity measure is applied to evaluate the distance between histograms of a target candidate and target template. Finally, a collaborative similarity measure is proposed from two different models and the corresponding likelihood of the target candidates is input into a particle filter framework to estimate the target state sequentially over time in visual tracking.●A multi-scale patch based appearance model with sparse representation is proposed and an efficient scheme involving the collaboration between multi-scale patches is established by sparse coding histogram. The key idea of the method is to model the appearance of an object by different scale patches, which are represented by sparse coding histograms with different scale dictionaries. Then a sparse coding-based similarity measure is applied to evaluate the distance between histograms of a target candidate and target template. Finally, the similarity score of the target candidates is input into a particle filter framework to estimate the target state sequentially over time in visual tracking. Additionally, to decrease the visual drift caused by occlusion, we present an occlusion handling strategy which takes spatial information of multi-scale patches and occlusion into account for robust tracking.