Video Annotation Model Based on Multi-Label Classifier and Fuzzy Knowledge Representation Schemes

M.Sumithra, V.Mercy Rajaselvi

Video annotation is a promising approach to facilitate video retrieval and it can avoid the intensive labor costs of pure manual annotation. But it frequently encounters several difficulties, such as insufficiency of training data and the curse of dimensionality. Video annotation is processed by three steps, In the first step, in order to extract the key feature, video is taken as input and a single frame is extracted from the video by using the video cutting tool. From the selected frame GIST Descriptors for spatial structure and the feature vector of 8x 8 encoding samples are extracted. In the second step, classifier are trained and annotation is done. In the training phase, trained classifier is obtained by using SVM algorithm and in the classification phase, labels are given to the object using the trained classifier.  In the third step, scenes are recognized by inference based algorithms which takes object labels as input. The inference based algorithms are used for annotation refinement and scene recognition. These algorithms use fuzzy knowledge representation scheme based on Fuzzy PetriNet and KRFPNs. KRFPNs is defined to enable reasoning with concepts which is useful for video annotation