To satisfy different positioning accuracy demand in the movement of the service robot, a hierarchical localization method based on vision was proposed considering the positioning cost and computational efficiency. The hierarchical positioning method could be divided into two parts: the localization method based on global vision and the localization method based on binocular vision. The global vision-based localization method was applied to calculate the initial coordinates of the robot and realize real-time rough positioning with the dead-reckoning method in the first stage. In the last stage, the binocular vision-based localization method was adopted to obtain higher positioning accuracy to make sure that the service robot can successfully grab the object. Experiments indicate that the proposed algorithm can accurately evaluate the positioning performance, obtain up to ±2 cm positioning accuracy, and absolutely meet the positioning requirements of the indoor service robot.
