Localization Of Mobile Robot With Omnidirectional Vision Under Known Environment

The intelligent migration robot already became a very important project in recent years in a robot research, along with the research development, robot's intellectualized level is getting higher and higher, the independent ability is also getting stronger and stronger. But mobile robot's localization is the very important direction in the intellectualized research, is precondition which the robot intellectualization realizes. The paper studied the question of the mobile robot carrying on the independent localization under the known inherent environment through the panoramic sensor.First, s formation model of the panoram sensor's image has been carried on the study, the SIFT(Scale-Invariant Feature Transform) algorithm for panoramic picture feature extraction is conducted in the research and the discussion. Consumes the major problem of huge match time in view of the classics SIFT algorithm when it is operand, a set of improvement feature extraction method has been designed, and this method is a solution of the real-time question, is able to improve the localization time efficiency.Next, a characteristic cartographic representation method which the characteristic chart and the grid chart unify has been designed. Using the index search's thought that the image characteristic according to the position, the category information carries on the preservation. Through this way can reduce the massive redundant information search and the inquiry in the match process while not reduce the characteristic information in the foundation,Once more, a set of unique map match search mode has been designed according to the distributed characteristic of panoram vision image. This reconnaissance method proposed the thought that the environment search and the direction approach with graduation search ,and may extricate the robot from the arduous map storage characteristic search, and can cause the robot to reduce the localization match time. And in carries on the analysis after the map survey the extremely quick hunting zone will concentrate in the current position in the region.Finally, Considering the possibly occurring of bad matches in the the image match localization process, a model of the robot state in motion is carried on. A method is carried on by returning the robot positional information that taking by ultrasonic sensor and the image match result, with the Kalman Filtering thought that has designed the position correction locate mode.