Temporal spatio-velocity transform and its applications

Object tracking is important in various applications such as video surveillance systems, video annotation systems, and human interaction classification systems. Occlusion and noise are the most significant problems in object tracking. In order to overcome these problems, I introduce the temporal spatio-velocity (TSV) transform, which extracts pixel velocities from image sequences. The TSV transform appends the velocity axes to the image sequences and separates occluding objects based on their velocities. The TSV transform is derived from the Hough transform over windowed spatio-temporal images. I present the methodology of the transform and its implementation in an iterative computational form. The intensity at each pixel in the TSV image represents a measure of the likelihood of occurrence of a pixel with instantaneous velocity in the current position. Binarization of the TSV image extracts blobs based on the similarity of velocity and position. The TSV transform provides an efficient way to remove noise by focusing on stable velocities, and constructs noise-free blobs. In this dissertation, I introduce three applications using the TSV transform. The applications are (i) human interaction recognition system, (ii) object tracking system in occluding environments, and (iii) soccer player tracking system. The human interaction recognition system uses side-view image sequences and tracks persons walking on sidewalks. Then it recognizes the interactions between two persons such as "two persons meet from different directions" and "one person follows another person". The system correctly tracks persons and recognizes the interactions between them. The object tracking system in occluding environments tracks moving objects behind static obstacles, such as trees and fences. Although the static obstacles divide moving objects into several pieces both temporally and spatially, the system correctly tracks the objects. The soccer player tracking system tracks soccer players and referees using the ordinary TV broadcasting images. Although the soccer players make complex movements and the camera moves frequently, the system correctly tracks the players and referees.