The application of multi-rotor UAV in vision has been developed rapidly in recent years, and multi-rotor unmanned aerial vehicles (UAV) can’t be widely used without the aid of a visual system. The application of visual technology in UAV has attracted more and more attention with the continuous development of image processing and environment reconstruction technology. Getting accurate depth map from the stereo image has always been the pursuit of researchers. Stereo matching is the most important part of stereo vision. An improved matching algorithm is proposed to solve the problems of high image noise and low image accuracy in binocular vision imaging of multi-rotor UAV after binocular information fusion. Experimental results show that the proposed algorithm has advantages in accuracy compared with other algorithms. The validity and reliability of the proposed stereo matching algorithm for multi-rotor UAV based on binocular vision are verified by indoor simulation tests and real environment tests of a UAV.