| Summary: | 碩士 === 國立臺北科技大學 === 資訊工程系研究所 === 103 === This paper presents the wearable visual aid system to improve the daily lives of the visually impaired inconvenience. Provide visually impaired persons &;quot;landmarks remind&;quot;, &;quot;traffic alert&;quot; and &;quot;Item Identification&;quot; wearable visual aids and the establishment of image information platform. White cane for the space above the ground providing insufficient information, insufficient number of global guide dogs, for the visually impaired unrecognizable shape of articles similar issues such as improving the design. Use voice reminder visually impaired landmarks information, articles and obstacles appear warning information to assist visually impaired persons &;quot;safety guide &;quot;,&;quot;control environment&;quot; wearable visual recognition system.
Use Xtion Pro implement, through color image and depth image information real time RGB-D video streaming analysis. Obstacle warning adopt Depth Segmentation, Edge Detection, Contour Extraction position appears to mark the obstacle to voice warning. Use Speeded-Up Robust Features algorithms to extract item and landmarks features, establishment Multiple Randomized Kd Trees characteristic index data matching feature matching algorithm using Approximate Nearest Neighbors. In this experiment, the landmark recognition rate of 85.87%, in item identification recognition rate of 86.16%, the obstacle warning detection rate of 95.67%.Visually impaired test indicates that the secondary has a wearable visual aid to improve their daily lives inconvenience, can help to understand the surrounding information, set up their own mind map.
|
|---|
