Human action recognition using optical flow analysis of video with static background

Abstract

Human Action Recognition (HAR) is an important and one of the many challenging areas where complex problems are targeted by computer vision researchers. Automatic recognition of human activities in video would be useful for surveillance, content-based summarization, and human-computer interaction applications. Still it remains the challenging problem in the area of Human Action Recognition. Many works have demonstrated the difficulty of the problem associated with the large variation of human action recognition. This research work incorporates a new approach of the LucasKanade (LK) method for Human Action Recognition. Background subtraction method is applied for separating the foreground objects from the static background image in a sequence of video frames. The preprocessed silhouette images are stored as training dataset which is used for action recognition and optical flow analysis. newline