Service robots want to achieve intelligent interaction, self-tracking is the foundation. At present, the tracking of indoor robots is mainly for human body tracking or object tracking with auxiliary markers, only one type of target can be tracked. In order to solve the above shortcomings, this paper designed an automatic omni-directional robot, using a common camera, according the different target using different strategies to achieve the tracking of any tracking, regardless of their status. The tracking strategy can be divided into two broad categories. One is tracking the human. Using based classifier face tracking and feature point tracking, Set the face tracking as a high priority of tracking, and the feature point tracking is added to ensure the tracking integrity when face blind spots appear. Another is the tracking of objects in any state. In this case, the feature points are used to track the target feature points, using Shi-Tomasi algorithm and Lucas-Kanade algorithm. The 2D coordinates of target sub-pixel level accuracy are obtained. Combined with laser range finder, build three-dimensional coordinate positioning. Finally, this paper verifies the feasibility of the system through experiments. The success rate of tracking is over 90%, which has high reliability.