Articulated human body parts detection based on cluster background subtraction and foreground matching

Detecting people or other articulated objects and localizing their body parts is a challenging computer vision problem as their movement is unpredictable under circumstances of partial and full occlusions. In this paper, a framework for human body parts tracking in video sequences using a self-adaptive cluster background subtraction (CBS) scheme is proposed based on a Gaussian mixture model (GMM) and foreground matching with rectangular pictorial structures. The efficiency of the designed human body parts tracking framework is illustrated over various real-world video sequences.