Goal Seeking System Based on Panoramic Vision Technology in Autonomous Mobile Robot

• Main Authors: Tzu-yuan Chang, 張子源
• Other Authors: Min Fan Ricky Lee
• Format: Others
• Language: en_US
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/dvuum4

碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 99 === The robots have been used in various fields, and they also have replaced people’s works which are dangerous. For instance, space adventure, mine detection, and even bomb disarmed. Therefore, the robots need more multi-function and precisely sensing. General robot use monocular camera to capture images within the camera view. Others use dual cameras to build stereo vision which detect the object distance. Those applications cannot sense all the environment information that is outside the camera view. As the result, we use PTZ (pan tilt zoom) camera to take several overlapping images in different directions. Then we stitch all the images together to form a 360 degree panorama image. By using panorama view, we can easily detect the objects that we want, and control the mobile robot to move toward the targets. This thesis uses highly distinctive feature descriptor (HDF) based on scale invariant feature transform (SIFT) to find the matching features between two overlapping images. Then eliminate the wrong matching pairs by RANSAC, and perform the cylindrical projection, image alignment and image blending. Finally we stitch the image and find out the objects by using ellipse fitting. Once we have the objects, the robot will touch the closest one.