Design and implementation of video summarization based on Hadoop

Design and implementation of video summarization based on Hadoop

South Peng Huacong

This paper intends to use cloud computing technology to transform the current dynamic video summarization technology based on motion process from single machine extraction mode into distributed mode. Therefore, in the face of data expansion, only by adding inexpensive computing nodes to improve the extraction rate of video abstracts, so that video summarization technology can be better put into the actual production. First of all, a lot of research on cloud computing technology and video summarization technology. This paper chooses the suitable open source cloud computing platform Hadoop distributed system, and analyzes the related mechanism of Hadoop distributed system and processing Video digest. In the aspect of video summarization, the general situation of video summarization technology is summarized, and the main process of dynamic video summarization based on motion process is introduced in detail. Secondly, this paper studies the development Rules and video processing technology of cloud application on Hadoop distributed system, points out the advantages of using Hadoop distributed system to deal with the video digest, and puts forward some problems such as mapreduce, frame integrality and motion event consistency of Video digest technology. Through the analysis of the processing process, the data storage process of Hadoop, expanding the block logic, simulating the motion events, taking the reasonable valve value, the solution is given, and the theoretical basis is prepared for constructing the video summarization system based on Hadoop. Thirdly, on the basis of the previous theoretical analysis, the demand analysis of the video summarization system based on Hadoop is designed, and the whole system is implemented. Finally, the experiment data show that when the data volume increases to 350MB, the two-node video summarization system based on Hadoop is more efficient than the single machine extraction mode. And with the increase of data, the efficiency of the video Digest system based on Hadoop is more and more high, and the final processing period tends to be 1.63 times times of the stand-alone mode, and has good expansibility.

Design and implementation of video summarization based on Hadoop