Spatio-Temporal Interest Points (STIP) Based Method Of Recognizing Human Action

Human action recognition is an active and challenging area of research in the field of computer vision. It can be applied in many applications, including video surveillance, intelligent vehicles, sport video event analysis and video retrieval. So there is important theoretical and practical value to carry out research on human action recognition.This thesis focus on the local Spatio-Temporal Interest Points (STIP) based method of recognizing human action. For the problems of human action recognition in complex scene, based on a comprehensive survey of the state-of-the-art of human action recognition, the major works and contributions are summarized as follows.1. Summarize existing human action recognition methods. Access to a large literature, existing human action recognition methods based on STIP are outlined. The widely used STIP detector and descriptor have been illustrated in detail.2. Implement visual codebook based on human action recognition Approach. Visual codebook based approach is the primary method of target recognition. First, using clustering algorithm to build visual codebook; and the video is represented as visual codebook feature; finally, use SVM to classify the video. Meanwhile, the work in this thesis compares the performance of different features in a complex dataset.3. Proposed motion difference flow based human action recognition method. Currently in recognizing human action, the motion descriptor is based on the optical flow estimation in video. In complex scene, due to camera motion, object motion and other reasons, the motion estimation has some deviation. To end this, motion difference flow is proposed to represent motion feature of STIP. Experimental results verify that the proposed method is more accurate than existing action recognition methods.