A study on obstacle detection for the visually impaired

• Main Authors: Jun-Zin Chen, 陳俊智
• Other Authors: Day-Fann Shen
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/8tsk73

碩士 === 國立雲林科技大學 === 電機工程系 === 105 === This thesis develops a system to detect obstacles for the visually impaired. The proposed system is composed of pre-processing, transform between color and depth information, and ROI detection. In the pre-processing, we use median filter and image inpainting to deal with some regions in the depth maps. We align the color images and depth maps based on the transform between color and depth information. In the ROI detection, a region growing algorithm and depth information are used to find the walkable region and then the obstacles in the walkable regions are detected. After obstacle detection, we also analyze the information of obstacles such as size, distance, and location. Here we build the proposed system based Kinect and evaluate the proposed system in indoor enviroments. Experimental results show that the proposed system can not only detect the walkable region but also locate obstacles well. Keywords: obstacle detection, Kinect, image inpainting, region growing.