Semi-supervised tensor-based graph embedding learning and its application to visual discriminant tracking

摘要

An appearance model adaptable to changes in object appearance is critical in visual object tracking. Inudthis paper, we treat an image patch as a 2-order tensor which preserves the original image structure. We designudtwo graphs for characterizing the intrinsic local geometrical structure of the tensor samples of the object and theudbackground. Graph embedding is used to reduce the dimensions of the tensors while preserving the structure ofudthe graphs. Then, a discriminant embedding space is constructed. We prove two propositions for finding theudtransformation matrices which are used to map the original tensor samples to the tensor-based graph embeddingudspace. In order to encode more discriminant information in the embedding space, we propose a transfer-learningbasedudsemi-supervised strategy to iteratively adjust the embedding space into which discriminative informationudobtained from earlier times is transferred. We apply the proposed semi-supervised tensor-based graphudembedding learning algorithm to visual tracking. The new tracking algorithm captures an object’s appearanceudcharacteristics during tracking and uses a particle filter to estimate the optimal object state. Experimental resultsudon the CVPR 2013 benchmark dataset demonstrate the effectiveness of the proposed tracking algorithm.