Video data mining based on K-Means algorithm for surveillance video

In this paper, we propose a new data mining algorithm, which is used in surveillance video of stationary places. The algorithm combines Background Subtraction with Symmetrical Differencing in order to extract moving targets. According to the amount of motions occurring in video frames, we divide the video into different segments. Video segments are clustered via the improved K-Means algorithm. Then we find the abnormal events, congestions and similar situation retrieval effectively in this way. To a certain extent, intelligent surveillance is implemented well.