Method for real time surface tracking in unstructured environments

Abstract

Although object surface or human face recognition systems have already achieved an impressive efficiency in recent years, the object surface or human face as well as object to object or person to person tracking task still remains challenging, especially under non-ideal conditions, as is common in driverless vehicles, surveillance or security based applications, particularly in unstructured environments. In this disclosure, data captured through artificial vision system, e.g. CCTV surveillance is analyzed identifying the object or person and object's or person's movements across the respective platform. Image regions are categorized based on the population density. The regions are subsequently decomposed adaptively into tree of image patches. Furthermore, the surfaces are later stored efficiently and retrieved as patches. Consequently, the method disclosed is an efficient and adaptive method of constructing a path tracking data structure of arbitrary surfaces with graphical information facilitating better computability.