Abstract
This paper describes a body feature extraction method from depth camera. The camera is used to collect point cloud data representing the human body shape. The edge point data along the human body contour are first identified in the coordinate system for feature extraction. The adjacent points on the contour curve are subsequently used to describe the shape of the human body. The horizontal coordinates of the points are extracted in increasing or decreasing sequence. An efficient feature extraction algorithm is proposed to analyze the coordinate values of these points using a sequence theorem to locate the body features. When there is a change between the two connected sequences (changing from an increasing sequence to a decreasing sequence or vice-versa), the turning point between the two sequences is defined as the extracted feature point. The feature points such as the armpit points and the crotch point are then found on the concave parts of the human body without pre-marking. The proposed method is tested on male subjects, with human body feature points successfully extracted. The extracted feature points can be further used to collect body dimensions for various applications.