Shopping scenarios semantic analysis in videos

This paper presents a computer vision system that analyzes a scene, recognizes human behavior, and produces a semantic interpretation of the recognized behavior. Scene analysis consists of detecting motion and tracking detected objects. Then, the system detects events that define the current behavior of a person. Once behaviors are classified the system recognizes various predefined scenarios. Finally sentences are generated to describe each person´s action in the scene. Our research is developed in order to build a realtime application that recognizes human behaviors while shopping. In particular, the system detects customer interests and interactions with various products of a point of purchase.