Human Identification Using Human Body Features Extraction

A system that can mark personnel attendance becomes an increasingly important tool. It usually requires human action in order to detect the presences of a person such as putting finger on a scanner in a fingerprint detection system. There are some systems that do not need human action (such as the use of camera in a face recognition system) but it requires human face database. This paper introduces human identification system implementing human body feature extraction, which can track the presence of persons in a room. The system uses camera as input device to remove the special action needed. Features on human body was chosen because they tend to be easier to detect and serve as a strong identity to mark person who enter a room, yet we do not need to train body features first. There are two main processes in this system. First, the system detects person entering a room. Second, it also detects the same person leaving the room. This application was implemented using image processing techniques such as human detection using HOG descriptors, HSV color conversion, and template matching. Tracking failures from this application could happen because this system still could not handle some special cases.