Deformable Object Tracking Using Clustering and Particle Filter

keywords: Visual object tracking, data clustering, object segmentation, cluster correspondence
Visual tracking of a deformable object is a challenging problem, as the target object frequently changes its attributes like shape, posture, color and so on. In this work, we propose a model-free tracker using clustering to track a target object which poses deformations and rotations. Clustering is applied to segment the tracked object into several independent components and the discriminative parts are tracked to locate the object. The proposed technique segments the target object into independent components using data clustering techniques and then tracks by finding corresponding clusters. Particle filters method is incorporated to improve the accuracy of the proposed technique. Experiments are carried out with several standard data sets, and results demonstrate comparable performance to the state-of-the-art visual tracking methods.
reference: Vol. 37, 2018, No. 3, pp. 717–736