Human Motion Recognition Using Optical Flow Based Particle Filtering
Human motion recognition has been a significant topic in computer vision due to its numerous applications such as video surveillance, human machine interaction and video retrieval. One core difficulty behind these applications is mechanically identifying low-level actions and high-level activities of concern. This paper presents a technique of automatically identifying the motion accomplished by humans. The speeded up robust features (Optical flow with kinetic energy algorithm) are made use of, for extraction of the key-points in both spatial and temporal domains. The Spatio-Temporal Difference-of-Gaussian (STDoG) pyramid is primarily developed which is further used to find the maxima and the minima points which provides the interest points. The key points are in the xy, xt and yt planes where xy matches with the spatial plane, xt and yt planes represent the temporal domains. Experiment is performed on a video, covering multiple action detection on KTH dataset.
Keywords: KTH, Spatio-Temporal Difference-of-Gaussian, speeded up robust features
- There are currently no refbacks.