A YOLOV5-PF-based Approach to Visitor’s Interest Discrimination applied to Pavilion Tour-guide Robots

In the field of human-computer interaction, accurate proactive interaction behaviors by robots can effectively attract user attention and enhance user experience. In this work, a method called the visitor’s interest discrimination method is proposed for exhibition and museum guide robots to identify visitors in the environment who are suitable for initiating proactive interactions before initiating human-computer interaction requests. The proposed method analyzes the bounding boxes of targets based on YOLOv5-PF (YOLOv5-Poo1Former) object detection and combines a weight analysis algorithm that integrates the interaction area and interaction distance between visitor and showcases. This enables the estimation of visitor attention towards the showcases and inferring visitors’ interests in the showcases. Through experimental validation, the proposed method accurately discriminates visitors’ interests in the environment.