Multi-robot visual support system by adaptive ROI selection based on gestalt perception

Abstract

This paper deals with a visual assistance system for a remote control of a working robot. While an operator controls the working robot, an another autonomous monitoring robot evaluates a suitable viewpoint to capture the working situation, and dynamically moves to the optimal viewpoint for monitoring. The monitoring robot detects a gestalt of the scene, where a latent and expected interactive objects are included in the region of interest (ROI). Latent Dirichlet Allocation (LDA), which is a probabilistic topic model, is employed to judge the degree of risk around the working robot. Such a situated ROI is transmitted to the operator in real time. Experimental results shows that the proposed system can enhance the remote control mission by reducing the occurrence of error significantly.