UAV Autonomous landing algorithm based on machine vision

Autonomous landing capability is very important for UAV, this paper presentes a UAV landing method based on visual navigation pose parameters calculateion. The navigation algorithm is based on monocular vision, and the shape and size parameters of land objects need to be carefully designed. The contour extraction method is combined with the Angle detection method to obtain 8 Angle points with good geometric distribution and moderate number. The algorithm can calculate the relative pose parameters of the UAV and the landing platform accurately by only using these 8 Angle points without the need of depth information. The 8 angular points extracted are directly used as the input of detection and tracking, which ensures the real-time performance of the algorithm and reduces the processing time of stable tracking landmarks. Finally, the real-time calculation of the flight pose parameters is carried out. The results show that the algorithm has an average period of 49.523 ms(about 20 fps) and can meet the real-time requirements of autonomous landing visual navigation when the speed is not higher than 2m/s.