Survey on Real-time Activity Detection and Recognition in Video

Abstract

Real-time object recognition and detection is the ability to automatically analyze object to recognize and assess temporal events which do not rely on a single image. It is the mechanism by which a video is stored, data gathered and data evaluated for the purpose of collecting domain specific knowledge. Object Identification is the method of identifying instances of real-objects. It enables several objects to be identified, focused, and detected within an image, picture, or in real time. The identification of anomalous events and artifacts through video becomes quite difficult owing to the uncertain existence of the phenomena, the background under which the incident took place, the absence of sufficient amount of anomalous ground truth testing data and other considerations correlated with weather variability, lighting conditions and the operating state of the cameras recorded. This paper aims to research and evaluate different anomalous behavior detection and event tracking strategies based on film. Various activity and object detection systems were provided the emphasis. The methods are contrasted from both precision driven activity detection viewpoints and real-time computation driven activity detection. This paper further focuses on work problems and obstacles, technology contexts, reviewed databases and potential operation and object detection directions.