Real Time Unusual Event Detection In Videos To Enhance Atm Security
In present-day applications, tracking object in low resolution video is a stimulating task due to loss of distinctive aspect in the visual outward look of moving target object. The present-day approaches are typically involved on the enhancement of Low Resolution (LR) video by super resolution techniques. However, these approaches need massive computational cost. This cost again rises if we are handling with event detection. A system has been proposed which is capable of detecting unusual events and for enhancement of ATMs security where simple low-resolution cameras are used because of their lower cost. This technique is able to detect the unusual events like face masking, camera masking, fight or overcrowding in the low-resolution video merely by using concepts such as Convolutional Neural Network. It processes low resolution clips and also sends alert messages to the concerned authorities in investigation (surveillance) department for enhancing the ATMs security and prevention of thefts.
Keywords: Video Processing, Feature Extraction, Decision Trees