Clock-Modeled Ternary Spatial Relations for Visual Scene Analysis

The analysis and the description of complex visual scenes characterized by the presence of many objects of interests involve reasoning on spatial relations such as “above”, “below”,“before”, “after” and “between”. In this context, we have defined these semantic concepts in terms of ternary spatial relations and we have formalized them using the clock model which is based on the clock-face division and the semantic notions of hours to describe relative spatial positions. The presented approach has been efficiently applied for the automated understanding of spatial relations between multiple objects in real-world computer vision image datasets.