Research Of Video Retrieval Base On Object Feature

With the fast development of Internet and multimedia technology, video is applied and developed widely due to its diversiform representation, abundant semantic content, and convenient ways of recording. As video information is heavy, the effective retrieval and analysis becomes a subject to be solved urgently.At present, the management and retrieval of video are mainly based on the description of video files, and it is not popular to operate the inside details of the files, in other words, we seldom retrieval video files based on their actual contents, which the users are more concerned about. In the thesis, a number of novel algorithms for video retrieval base on object feature are presented based on our researches and experiments. Our work contributes to the following aspects:In detection of moving vehicles, we lay an emphasis on the research of moving target detection in traffic video, and propose a fast background modeling algorithm. The experiments showed that using of these algorithms is comprehensive and effective.In extracting and analyzing the features of vehicles, we analyze and summarize the features of moving vehicles in the video, and build mathematic model to describe them. Besides, we give a strategy to extract the features of objects in the video.In researching on video retrieval algorithm, we propose a method of video retrieval base on the features of object and design a traffic video retrieval system, which can acquire the features of vehicles such as color, size, the time of getting in and out, etc. The experiments showed that we can get comparative exact information of the vehicles we need.