Based On Monocular Vision Research Of Auxiliary Unmanned Helicopter Landing

Mini unmanned helicopter which features low cost and high flexibility, has comprehensive applications in military and civilian fields. As an important part of the mini unmanned helicopter, visual system has been paid great attention by many colleges and research organizations all over the world. This dissertation proposes a monocular method, which can detect ground area and posture parameters in unconstraint environment when an unmanned helicopter autonomous landing with no artificial target.Firstly, in this dissertation, we introduced the background and the content of our work. In addition, the composition of autonomous landing system and previous methods are elaborated. The advantages and disadvantages of these methods are discussed. The dissertation described the camera models and camera calibration. Harris corners and SIFT feature extraction are described in detail, we compared the difference of these two methods and their scope of application.Secondly, this dissertation studied how to detect ground area for helicopter autonomous landing. The fundamental matrix and homography matrix induced by ground plane between the views are computed, the essential matrix is derived from fundamental matrix for estimating camera motion parameters, feature points’3D positions are computed up to scale, and normal vector of plane is estimated. When the ground slop is acceptable, homography mapping image points in the second picture to the first picture, then classify the image into ground areas and non-ground areas based on whether points overlap. The experiment results indicate that this algorithm detects the ground areas is available for landing an unmanned helicopter reliably.In the end of this dissertation we studied how to estimate the position and attitude of the helicopter in the landing process by the homography decomposition method, compared the estimation results with the value calculated by control points method. The experimental results illustrates that the precision of the position and attitude of presented algorithm can meet the requirements of unmanned helicopter autonomous landing.