Automated inspection of door parts based on fuzzy recognition system

The article presents a comprehensive strategy of door parts (locks, handles, doorplates etc.) examination as a paradigm of active sensing. It covers the whole process — from segmentation, through initial hypothesis generation based on fuzzy inference, to final recognition and precise localization of the keyholes in a robot base coordinate system. The strategy is a preliminary stage of a door locks opening procedure. The image analysis process is divided into three steps. First, an initial region of interest is localized using a RGB-D low resolution camera mounted on the robot's head. It is then categorized, based on its properties, as a lock, handle, doorplate etc. Finally it is inspected using 2D camera mounted on the robot's arm. Thanks to the preliminary localization with the head camera the robot can look at the point of interest at sight, without time-consuming thorough inspection of the whole door. The whole system was formally specified as an embodied agent. For that purpose a system modeling language was used to specify embodied agent subsystems behaviors. The proposed strategy is verified experimentally using Velma prototype service robot.