A Geometric Localization Method with Monocular Vision for Robots

Robot soccer is a classical problem in artificial intelligence. In robot soccer competitions, how to locate the robot properly is a key problem. This paper proposes a geometric localization method with monocular vision for robots. Firstly, the image is segmented and attributes are obtained by converting color space and RLE algorithm. Then a monocular vision localization model is built, in which both the pitch angle and direction angle of the head are not equivalent to zero. After that, we calculate the target coordinates and the corresponding pitch angle and direction angel by rotating the image plate, obtaining the relative position of the target. The simulation result shows that the proposed method reach high accuracy in localization.