Tracking torso of a human body in real time from a single general-purpose camera
Ke-Yu Chen, Class project (EE576) 2011
Human 3D modeling and reconstruction from 2D sequences has been researcher’s interests for years. Torso is the main part of the human body and connects to our limbs (through shoulder joints) and legs (through pelvic joints). Finding the accurate position of torso is important in human 3D modeling and reconstruction. In recent years, many products such as Kinnect provide excellent performance of human 3D modeling and reconstruction by multiple cameras. However, when high‐tech devices are not accessible or old videos recorded by a single camera are the only data source, we need to explore this kind of reconstruction process from 2D sequences purely captured by a single camera. For this purpose, I designed an algorithm to track the torso of a human body in real time from a single general‐purpose camera. The algorithm exploits silhouette of the human body and extracts the torso with distance transform and dynamic threshold techniques.