Recently, in the field of video analysis and retrieval Human action recognition in video is an important research and challenging topic. An efficient video retrieval is needed to search most similar and relevant video contents of a large set of video clips. Many methods have been used for the efficient retrieval of videos. In this Adaptive weighted pyramid matching kernel (AWPM) has been used for efficiently retrieving videos by recognizing human actions in realistic videos. This can be done based on a multi channel bag of words which is constructed from local spatialtemporal features of video clips. AWPM is the extension of spatialtemporal pyramid matching (STPM) kernel leverages in spatiotemporal granularity level and in multiple feature descriptor types to build a suitable similarity metric between two video clips. STPM uses predefined and fixed weights and hence the proposed matching algorithm estimates adopts channel of weights based on the Kernel target alignment of training data. The following work is analysis over the content based video retrieval in large database through various mechanisms available in the literature.
展开▼
