Video Streaming Media Bigdata Study Based On Semantic Analysis Process

Video semantic analysis is the description and logical representation of the contents in video information, which is a complicated process of information processing in many research areas. BigData technology has the advantage of data storage and data computing, and granular computing theory is good at data representation and feature classification. Therefore, both BigData technology and granular computing theory are combined and applied to video semantic analysis system, which can solve the key problems such as data storage, computing, and representation.It is focused on structuring the storage framework for the big data in the streaming media, building the analysis model of video semantic, and proposing the key algorithms for video analysis. Firstly, the key problems in video data processing are discussed and a novel big data storage framework for streaming media is proposed to solve the storage and inquiry of the video semantic analysis results. The design solution is to storage and retrieves video data by time features. Secondly, the BigData technique is employed video semantic analysis and the granular computing theory is applied to the structure description of video data. On the basis, a model for the hierarchical video semantic analysis is built under the different granularities. At last, the detection for the moving object and shadow in video were discussed, and the motion vector detection-based shadow suppression algorithm was proposed and implemented in the big data systems.The main research contents and innovative works in the dissertation include:1. Structuring a novel storage framework for the addressing, retrieval and analysis of big data in the streaming media by time features. This framework supports the compression of encoding and decoding of streaming media data. It can store and address by frames and can realize the quick positioning and unified storage of analysis results in the mass storage medium.2. Structuring the metadata description framework of semantic analysis and setting the interface for hierarchical database. A multipurpose database for the big data storage and retrieval in streaming media is designed to build their respective metadata model and database interfaces for different applications. It can solve the connection problems in the heterogeneous video capture modules and subsystems.3. Mapping image features involving image objects, video objects and semantic objects into granules and building the granular model for video streaming media containing granular property structure. The video granule in different granular layer can be treated. It simplifies the algorithm structure of video analysis and good for the realization for parallel computing by adopting the Bigdata technique.4. Proposing the motion vector detection-based shadow suppression algorithm. The motion vector is extracted from the video coding information and applied in the shadow suppression of the moving objects. The algorithm used video streaming media granular model for video background modeling which obtains a good result of shadow detection.