Analysis of Low and High Level Feature Extraction in Human Movement

Abstract

Segmentation of image is very challenging in tracking down the movement of humans. In this paper we applied optical flow to differentiate the features between walk and run. The features extraction generated was based on shape and velocity of the movement. The features based on shape will generate perimeter, area and centroid meanwhile the usage of velocity will generate angle and magnitude. The implementation of these techniques will produce the understanding of intensity in term of high and low level feature extraction. The output shows that the dynamic movement of human can be evaluated consistently by shape in the situation of constant lighting but the angle distribution or magnitudes are more robust in any environment