ABNORMAL EVENT DETECTION IN VIDEO USING FRAME SUBTRACTION AND FEATURE BASED TRACKING METHOD

Abstract

Nowadays, there is large requirement of video surveillance system in security and intelligent system for the public and finding out motion of real time object is a major task in video surveillance system. A single surveillance camera can record all activity in day and night. In this paper, we used an approach to find out abnormal activity based on features of object which is independent of object size, shape and speed. To detect the suspicious moving object of abnormal behavior, we have used foreground subtraction approach and features based descriptor. In frame subtraction, we will calculate the difference of active pixel of successive frame in motion and find out the region of interest by setting a fixed value of threshold. We will compare the mean of illumination of active pixel of frame in motion by threshold value, and locate the suspicious object if its value is more than the threshold value by setting pixel value 1 and non-interest portion with pixel value 0. In our approach, we only focus on the frame in motion having active pixel, so for that we find the frame in motion to increase efficiency. For this purpose, calculate the peak values of change in illumination of active pixel of initial 10% frame. If the difference of illumination in successive frame is more than that peak value then we can say that there is a high probability of change in motion and we count it as “frame in motion” otherwise ignore the frame. Advantage of this approach is that we can save time and also increase efficiency. For track and locate the object, we use features based tracking method. The proposed algorithm requires less computation by using efficient approach and also increase accuracy in video surveillance system to detect abnormal activity in real time environment.