Research And Implementation Of The Monitoring System Supporting Video Retrieval On Mobile

A research and implementation of the monitoring system supporting video retrieval on mobile was conducted in this thesis. The users can not only obtain the real-time monitoring scenes of server-side via mobile, but also complete the remote retrieval of the past video data and achieve the corresponding retrieval results. In determining abnormal situation, the system will automatically send alarm information to the mobile terminal, thus solving the problem that the real-time personnel's supervision is needed for traditional video monitoring. Meanwhile, the server-side of the system is capable of intelligent analysis of monitoring data. It can store valuable shots for people without the meaningless background frame, so as to solve the problems of high pressure on the space and high redundancy of the saved video data.Firstly, a careful study of the related technologies in this system and an analysis of the advantages and defects of the current technologies are conducted in the thesis. Then the related design of shot segmentation algorithm and key frame extraction algorithm combined with the applied scenes in this system is carried out, overcoming some practical problems which are difficult for the current technologies and achieving good results. The innovative points of this thesis are mainly in the following three aspects:(1) The development of the function of video retrieval on mobile makes it possible for the place-to-place mobile video monitoring and retrieval, satisfying the users' demand for retrieval and view of historical monitoring video data;(2) A real-time shot segmentation algorithm which is suitable for motion detection of indoor fixed camera is put forward. The algorithm integrates the advantages of background difference with those of frame difference, solving the problems of the impact on the shot segmentation because of the background light changes in the monitoring and small motion in the video event and so on;(3) It comes up with a real-time key frame extraction algorithm which is suitable for motion detection of indoor fixed camera. This algorithm have dealt with the unstable extraction and poor adaptability resulting from the fixed threshold in the traditional key frame extraction algorithm with the background frame difference as the feature of the selected key frames, and the dynamically growing threshold as the threshold of the selected key frames. Meanwhile, this algorithm has less computational complexity, and more typical extracted key frames.Experimental results show that the main functions such as intelligent storage, remote alarm on mobile, remote video retrieval on mobile can be well implemented in the system. Compared with the traditional video monitoring system, in addition to reduction of the labor cost, improvement of the monitoring efficiency and the convenient and flexible use, this system is also able to expand the scope of monitoring and the function of mobile monitoring, giving the users a full control of the monitoring, which brings an extensive application value. What's more, compared to some previous algorithms, the shot segmentation algorithms in this thesis can better adapt to the light changes in the monitoring and small motion in the video event and so on, while the key frame extraction algorithm is able to overcome unstable extraction and poor adaptability resulting from the fixed threshold, which has the theoretical significance.