Approach to the Segmentation of Buttons from an Elevator Inside Door Image

For a robot, even though its navigation in indoor environments has been studied well, its movement between different floors is still a challenging topic since the robot should possess the ability to recognize and control the buttons on the elevator control panel for taking the elevator. In this paper, an automatic approach for segmenting buttons from an elevator inside door (EID) image is presented before pursuing the goal of recognition and control. Due to the vary styles of buttons being used onto elevator control panels and the reflection phenomenon existing in an EID image, it is not easy to deal with the button segmentation. To overcome this problem, based on the edge information of the EID image, a kernel function called projection-and-checking (PAC) and some refining processes are developed for the button segmentation, which will be useful for the robot vision application. Our experiments confirm the feasibility of the proposed approach.