Elevator button recognition using auto-slant correction and projection histogram

Abstract

An effective elevator button recognition method is proposed for a robot to go up and down floor through the elevator autonomously based on auto-slant correction and projection histogram. In the proposed method, we train a particular elevator button image only one time. In the process of training, firstly, we assign a particular elevator with a unique number (ID code), and then mark the elevator button area manually in the image captured from the particular elevator; secondly, the horizontal and vertical projection histograms can be obtained from the preprocessing result of elevator button area, the preprocessing includes illumination uniformity, binarization, morphological filter and so on; finally, the position of each elevator button is calculated according to projection histograms, and then the elevator button template is calculated by using elevator buttons' position. In the process of testing, first of all, the system judge if the elevator image is slant, the slant elevator image is corrected automatically according to preprocessing result of elevator image and then the horizontal and vertical projection histograms are obtained from the corrected elevator image; secondly, the precise elevator button area is confirmed by calculating the correlation of projection histograms received in the process of training and testing; finally, the position of each elevator button can be calculated correctly from precise elevator button area. The experiment results indicate that the elevator buttons can be recognized precisely in real time.