Robot Self-positioning Method Based On Omnidirectional Vision System,

The thesis designs and implements a novel omni-vision system for the application in the middle-size league of RoboCup– The World Cup of Soccer Robots, and does research on the robot's self-localization methods based on the omni-vision system. The omni-vision system can capture the rather perfect panoramic image realtimely, process the image fast and effectively, including image segmentation by color, extraction of features and detection of the white mark lines of the soccer field, and complete the robot's recognization of target. The robot can gain the accurate localization value consisting of position and orientation in the world coordinate system using the robot's self-localization method based on the information of mark lines and the Monte Carlo localization method the thesis designs.The main work of the thesis is as follows: we design a novel omni-mirror more applicable in RoboCup, which is made up of an isomeric horizontal mirror and an isomeric vertical mirror and can make the resolution of the imaging of the objects near the robot on the field constant and make the distortion of the imaging of the objects far from the robot small in vertical direction; we select a 1394 digital color camera and complete the development of its software program for data acquisition; we design a image processing algorithm for the panoramic image, which can segment the image by color, extract the features of image fast and effectively and complete the recognization of target; according the character of the imaging, we design a new fast Hough Transform algorithm for line detection, which can detect the white mark lines of the field realtimely, and then design a robot's self-localization method based the mark lines, which only uses the information of omni-vision system; finally we design a Monte Carlo localization method based on the information of omni-vision system and odometry, and present the localization results of two localization methods and analyze the experiment results and get the conclusions.