Research On Indoor Location Method Of Rotor UAV Based On Vision

With the rapid development of unmanned aerial vehicle(UAV)technology,UAV has been widely applied in logistics,patrol inspection,reconnaissance and other industries,and gradually expanded from outdoor application to indoor application.Reliable location and navigation technology is the prerequisite for the normal operation of UAV.Normally,UAV mainly relies on GPS inertial integrated navigation.However,in the environment of jungle,urban buildings and indoor,it is difficult or unstable to locate the UAV due to the interference or blocking of GPS signals.Therefore,vision,a low-cost and convenient location method,is often adopted in indoor applications.The problem is that it is easy to be affected by environmental conditions such as illumination and has poor stability,which directly affects the location accuracy of UAV and the robustness of the location system.Therefore,it is necessary to carry out research on the indoor location method of UAV based on vision.In this paper,the following research is carried out on the indoor location of rotor UAV.Firstly,aiming at the problem of large computation amount of traditional feature extraction methods and poor feature point extraction effect under special environment,the feature extraction algorithm of ORB is improved through image area segmentation and adaptive threshold method,which is verified by experiments.At the same time,LK pyramid optical flow algorithm is adopted to solve the problem that the optical flow algorithm fails when the UAV has a large movement.Secondly,in order to reduce the influence of illumination variation on the optical flow algorithm,the feature points of the high-frequency part of the image are extracted through gaussian filtering,and the PROSAC algorithm is used to eliminate the mismatched point of the optical flow.The experimental results show that the proposed algorithm has good anti-interference.Then,based on the idea of Meanshift algorithm,the optimal solution of optical flow is obtained,which further improves the estimation accuracy of optical flow.On the basis of the above work,the location scheme based on vision and IMU fusion is adopted in this paper.Through the method of angular velocity attitude compensation based on IMU,the velocity estimation deviation caused by attitude change is eliminated,and the abnormal points of velocity are eliminated.In addition,visual and IMU data are integrated to improve the location accuracy and robustness of the UAV location system.Finally,a vision-based UAV location system platform is built,and the auto-calibrated PC software is developed to verify the above methods in the actual flight scenario.The experimental results prove that the proposed method can replace the GPS to provide information such as position and velocity for the UAV.