Small-size Unmanned Aerial Vehicles (UAVs) have found an increasing application in GPS-denied environment.Stereo vision can be used for UAV target detecting, tracking and locating in complex environments.The thesis presents the techniques of binocular camera calibration, feature point detection, extraction and matching, binocular measurement and so on.It is implemented to track the specified target in image while realizing target positioning, with final calibration error within 0.2 pixels and positioning error within 10 cm.We made verification to the algorithm by taking a ground robot as the target to be tracked, and found that it can achieve a fine tracking result when the target has short disappearance time.