Abstract-Three-dimensional reconstruction is used to get more detailed information on the current scenes such as the overlook of a city or a car. The process of reconstruction is based on binocular vision system. In this paper we apply binocular vision to quadrotors. As a result, more information is obtained in reconstruction when two quadrotors with cameras are searching in different circumstances. Video stabilization is used to guarantee the quality of images. Disparity map is computed by the semi-global stereo matching algorithm. The final reconstruction proves the proposed method to be effective.

Abstract-This paper presents a solution to trapped personnel detection and evacuation guidance in indoor fire scene based on quadrotor UAV. Our approach purely relies on UAV's own monocular camera as the main sensor, and therefore UAV can search and rescue in unknown, GPS-denied fire environments. In this paper, an effective method was proposed, which separated the mission into two parts: one was trapped personnel detection by using face detection algorithm, the other one was evacuation guidance, which was actually scale-aware navigation and path planning by using visual SLAM system. Costly computations were carried out on an external laptop that communicated with the quadrotor over wireless. The solution can be used directly with a Parrot AR.Drone quadrotor.