Scene Understanding: A Survey to See the World at a Single Glance

Abstract

Humans are extremely proficient at visually perceiving natural scenes and understanding high level scene structures. In recent times, scene understanding is a challenging and most important problem in computer vision. Images are visual, however the visual information can be with various features like shape, edges, texture and color. The main objective behind object detection is to identify what are all objects present in an image and where they all located. Understanding a scene integrates meaningful information at multiple levels to extract semantic relationships and patterns. The interaction between various objects is most intuitive and natural to human being. As compared with object recognition, scene understanding identify the target of objects and also the distribution of targets in a scene. Scene understanding have a significant impact in computer vision to perceive, analyze, and interpret the visual scenes which leads to new research areas. In this paper, we review the scene understanding concept including feature extraction, classification and scene recognition and related datasets. At last, we have concluded some challenges of object detection remain unsolved and can be scope for further research work.