Automatic Elevator Button Recognition with Convolutional Neural Network Feedback System Enhancement

• Main Authors: Yu-Ching Hsu, 許予晴
• Other Authors: Yung-Sheng Chen
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/w35gbm

碩士 === 元智大學 === 電機工程學系甲組 === 107 === With the acceleration of world economy circulation, the demand for automatic robot transportation grows immensely. In order to deal with the trend, the technique of image processing in computer vision is an inevitable key to empower robot the ability of cognition and understand the world we see. The application such as package delivering, clerical assistance, etc., are inseparable from the robots indoors navigation ability, including horizontal mobility and vertical mobility. Due to the less study focused on robot vertical navigation, this paper introduced a method to automatically locate (with 95.8\% accuracy) and recognize (with 100\% accuracy) the elevator indoor button panel with 32 panels in different patterns, different light conditions and different materials. And hope to become a force that fuels the robot ability for vertical navigation under an indoor environment.