Occlusion Handling In 2D and 3D Object Recognition

Abstract

Computer vision is the science of endowing computers or othermachines with vision, or the ability to see. It is concerned with the extraction,analysis and understanding of useful information from an individual image ora set of images. Object recognition is one of the fundamental tasks incomputer vision and is the process of finding or identifying instances ofobjects in digital images or videos. These could be, for example, humans, newlineanimals and static or moving objects.Object recognition methods frequently use extracted features andlearning algorithms to recognize instances of an object or images belonging toan object category. Variations in lighting conditions, viewpoint, background newlineas well as occlusion where the object is not fully visible can change theobserved image immensely. Many of these existing algorithms work well inconstrained scenarios with specific assumptions regarding object appearance,background color intensity information, duration of time for which an objectexists in the scene and/or limitation regarding number of objects within thescene. newline newline