Matching Characteristics of Two-Dimensional Images in Order to Determine the Three-Dimensional Position from Them for Robots

The basic idea used in this thesis is use of natural features as the environmental symptoms for positioning. There are four basic steps for positioning. The first step involves the extraction of natural features contained in the scene as environmental symptoms, the second step is finding correspondence between the found features in the database images and the acquired image by the robot, the third step is finding accordance by using number of found corresponding and the final stage is Robot positioning by using of found correspondences in the discussed scene. Performed innovation in this thesis is related to steps two and three that we will attempt to amendment and eliminate the correspondences with different methods after finding an initial correspondence. Three methods will be discussed at this stage which will be fully described. Finally, the study shows that the proposed method has strong points such as, independence on parameters of the SIFT operator or the ability to images with little correct correspondences which ultimately will increase the efficiency of the technique and improve program output.