Abstract: In order to realize the adaptive following control of two-wheel self-balancing robot in a simpler and more effective way, the Apriltag algorithm of monocular vision target recognition can passively solve the relative position of camera in the three-dimensional coordinates, thereby positioning the target. In the tracking and selection of target recognition, the discriminant method, the threshold method and the sliding window method are used to define, and the polynomial curve fitting method is used to track and predict the next state of the target movement, and the motion trend is determined. In the case of losing the target, according to the design of a set of search programs, the compensation follows and regains the target. Tests show that the method proposed in this paper has a significant effect in regaining the target after target recognition and loss.

Key words: monocular vision, autonomous following, target recognition, control strategy