Research On UAV Landing Attitude Detection And Estimation Algorithm Based On Monocular Vision

In recent years,drone technology has developed rapidly,and it has been widely used in service applications in cities.The acquisition of autonomous pose information of autonomous drones and autonomous landing are prerequisites for the application of urban services.With the rapid development of image processing technology and optical camera technology at the present stage,machine vision navigation landing technology has become a very hot research field,especially In urban environments where GPS signals are prone to interference,visual navigation landing techniques are particularly important.Therefore,based on the background of autonomous landing of UAV in the city,Taking the drone autonomous landing in the city as background,an estimation algorithm for landing attitude detection based on monocular visual drone was proposed in the paper.The algorithm combined with the landing motion model,using the information collected by the camera as measurement information,designed the motion estimation scheme based on the modified Kalman filter for the pose state of the UAV.This article completes the main research work as follows:(1)For the problem of camera distortion,the camera calibration technique was studied and the nonlinear problem was transformed into a linear problem.(2)In order to solve the real-time nature of UAV pose estimation,a fast template matching and segmentation algorithm is studied to segment the cooperative object with feature information.(3)For the problem of how to extract feature point information more accurately,A method of binarizing HSV color space is studied in this paper.After processing the image,the Canny edge detection algorithm is used to obtain the image edge information,and the contour extraction is performed.Corner detection,corner detection uses the Harris corner detection method.(4)First,a K-means clustering algorithm is designed to achieve fast extraction of 7feature points on the cooperative target image with high precision.Then a fast feature pointmatching algorithm is designed to achieve the world The coordinate system is matched.(5)In order to solve the problem of solving pose parameters of UAV,A pose estimation algorithm based on a perspective projection model is studied in this paper.and uses orthogonal iteration to optimize the estimation of the estimation results.(6)The modified Kalman filter and the monocular visual pose solution is used in this paper to combine the falling motion estimation scheme.The visual processed information is used as measurement information and the modified Kalman filter is used to achieve the next moment of pose.Prediction.The simulation results show that the pose accuracy is superior to the pure visual pose calculation method under the same noise intensity.