Semantic Content Extraction in Video Surveillance System

Recent increase in the video monitoring and surveillance has urged users to use video based application for extracting content and video semantics from it. The Results of such video monitoring and surveillance can be useful in various ways. Semantic analysis of videos has proven its worth in fields such as sports, national border security, suspicious activity monitoring along the LOC, etc. For fulfilling user's requirement such as an expected outcome in less time period and proper result with minimum storage space, the raw data and low level feature of videos are not sufficient. Though manual technique can prove handy in this regards, they are inefficient costly and time consuming. Increased demand for the video-based applications has revealed the need for extracting the content in video. We require automated system which moderate can surveillance and gives us information about important events happening in the premises. There are 3 layers of video data 1) Low level data (raw data) 2) Middle level data (content information) 3) High level Data (event information). The raw data contains the low level features of video like number of frames, length, pixel number, etc. these low level features are of very little or no importance to the user. The content, information consist information about object and occurrences of that object in video. The proposed system accepts videos as inputs which are divided into frames. Each frame is assigned to a particular number that gives event of frame in specific image from input video. These frames can be used for finding the position and speed of objects. Frame number plays the vital role in any video monitoring system.