The Method Research Of Semantic-distilling In Content-based News Video Retrieval

Currently, CBVR mostly makes use of the basic character of video and multi-vision character to realize combine searches. The shortcoming is that the technology can not using the semantic availability. How to define the semantic and distill the semantic information is the key-problem. So, the paper studies the semantic distilling technology in CBVR, and mostly discusses semantic distilling on news video and the analysis system design.First, the paper analyses the character of video data, and put forward the semantic model of video and the technology of semantic distilling. Then, the paper analyses the mostly arithmetic of key-frame distilling, and put forward a method of key frame distilling based on similar ratio between frames, and put forward another method to distill the key-frame. Later, a new algorithm is presented to detect and locate news local-captions using spatio-temporal distribution feature. In the TV symbol recognize arithmetic, the arithmetic firstly make inter-frame differential for the key frame, and divide up the image area of the TV symbol. The paper picks-up the wavelet character as the parameter of the TV symbol. At last, the method recognizes the TV symbol using the SVM arithmetic.At last, the paper put forward the character semantic distilling method,the sensibility semantic distilling method and the scene semantic distilling method. In the method of scene semantic distilling, the arithmetic divides up the news scene using the character of key-frame,the lens of toastmaster and the subject title. Then, the method distills the parameter of the scene and compared to the repository to obtain the scene semantic. In the method of sensibility semantic distilling, the arithmetic establishes the model of sensibility-space, and realizes the mapping between character and sensibility-space using the method of Probabilistic Neural Network (PNN). At last, the paper obtained the sensibility semantic using the repository of sensibility semantic.In the end, the paper realizes the news video retrieval system based on the arithmetic of semantic distilling.