Dynamic Objects Detection and Tracking from Videos for Surveillance Applications

Abstract

Artificial intelligence is a recent topic in computer science study right now. A machine must be aware of its surroundings in order to function intelligently. Humans require visual information to function. The field of research known as “Computer Vision” focuses on automatically interpreting images and image sequences. In other words computer vision is the scientific discipline that aims for extracting information from visual data. In a video sequence, moving object tracking refers to constantly recognising an object's position and orientation despite camera or object movement. The task of following one or more things around a scene from when they first appear to when they leave is known as object tracking. An item can be anything of interest in the scene that can be detected, and it is determined by the scenario's requirements. Object tracking systems are generally used in surveillance applications where moving individuals and/or vehicles must be monitored. Moving object detection, object tracking, and event recognition are the three basic components of any surveillance system. The performance of such system considerably depends on its first step that is detection of the foreground objects. These are the objects which are not background objects. These foreground objects have a important role in subsequent processes, such as event detection. To work properly, every object detector requires an object model, which specifies the look of objects, or a set of rules to determine whether a specific patch (region) from an input image contains an object of interest.