Vision Based Navigation For Unmanned Aerial Vehicle Autonomous Landing

Because of its unique advantages,UAVs are widely used in military and civilian applications.Especially in recent years,with the development of unmanned rotorcraft,UAV technology has become a hot topic.The autonomous landing of unmanned aerial vehicle(UAV)is an important stage in the flight process of unmanned aerial vehicle(UAV).Whether the UAV can achieve accurate self landing determines whether the task can be successfully completed.Due to their limitations,traditional navigation methods such as satellite navigation and inertial navigation can not meet the requirements of UAV autonomous landing.Therefore,based on the application background of unmanned vehicle autonomous landing,the autonomous landing navigation technology of UAV Based on vision is studied deeply in this paper.First of all,this paper introduces the research status at home and abroad,and summarizes and analyzes the future trends on the basis of learning and summarizing the existing research results.Combined with the previous research and results,a suitable landing landmark was designed.The landing landmark was built with a large map and nested small icons to meet the different heights of landmark recognition and pose parameter calculation.Then,the digital image processing technology is deeply studied.Combining with the landmark designed in this paper,the digital image processing method of the visual navigation algorithm used in this paper is given.After the image preprocessing of the navigation system has been done,a landmark identification and detection method based on the designed landmarks is proposed.Based on the classical aperture imaging model,the camera model and the related navigation coordinate system of the visual navigation system and the conversion relationship among the navigation coordinate systems are analyzed and studied.Then,Based on Zhang Zhengyou's plane calibration method,the camera is calibrated and the camera parameters are obtained.On this basis,an estimation method of visual pose parameter based on coplanar four points is proposed.The algorithm uses the Harris corner detection algorithm to derive the landmark image feature points and pair with the landmark space points to obtain the original point set containing the error.Then,The efficient set of points is obtained by using RANSAC algorithm.Using the obtained effective point set,the linear pose solution method is used to solve the homography matrix and the extrinsic matrix.And take this result as the initial value,the nonlinear optimization of Lenenberg-Marquardt is carried out to obtain the exact values.Finally,Based on Viusal Studio 2010 and C++ language we develop a set of visual navigation processing software.The software captures the camera image in real time and calculates the relative pose of the camera and landmark by using the pose parameter calculation method proposed in this paper.Then navigation error test and analysis test are carried out by using the navigation processing software and the three axis simulation platform.Test shows that at the height of 120 CM,position error using small icon recognition is about 5CM,the attitude error is about 4 degrees and can satisfy the engineering request.At higher altitude,the method of identifying large icons is used to ensure the feasibility of autonomous landing navigation algorithms at different heights.