Abstract In real life applications, identification of unusual events in low resolution video is a challenging task because of fact that there is loss of discriminative detail in the visual appearance of moving object. The current techniques are generally based on the upgrade of LR (low resolution) video by super resolution technique. These strategies require high computational expense. We present a design which can recognize unusual event such as weapon, face covered with helmet or multiple person detection without such kind of transformation and appropriate for upgrade of safety of ATMs where conventional low-resolution cameras are generally used due to their low fee/cost.
In proposed system we have used two techniques to detect the motion and image of particular object. Open CV algorithm is used for motion and Haar cascade is used for image recognition. These techniques have high accuracy and speed of operation. It has been analysed with the help of raspberry pi. This proposed method is applied to enhance the ATM security. Algorithm utilizes rolling average background subtraction method to identify foreground object from dynamic background in a scene. Our proposed system can observe the occurrence of unusual events in low- resolution video simply by using statistical property, standard deviation of moving objects and also send alert message to the authorized person.

Keywords: Object Tracking, Unusual event, Background subtraction, ATM security.

| DOI: 10.17148/IJIREEICE.2021.9708