Vision Aided Navigation For Unmanned Aerial Vehicle Landing On Ship

With the development of science and technology, unmanned aerial vehicle(UAV) plays a more and more irreplaceable role in aspects such as military and civilian. In military aspect it can reconnoitre,combat and gather information; In civilian applications are mainly involved searching and rescue work in dangerous environment. Make sure that unmanned aerial vehicle can land safely is the key of the unmanned aerial vehicle flight mission, the navigation method which based on computer vision technology is one of the most advanced navigation methods at present. Compared with the previous ways of navigation, method based on visual navigation not only can save a lot of manpower material resources, but also has low cost, high accuracy, high anti-jamming etc advantages; it has been a hot spot researched by a lot of scholars and institutions. In this paper, we study the vision aided navigation for unmanned aerial vehicle landing on ship, in order to improve the accuracy of pose parameters we proposed data fusion algorithm. And through the simulation experiment we verify the algorithm.First of all,we introduce steps of pose estimation algorithm based on computer vision technology for unmanned aerial vehicle in the process of landing on ship. firstly,the onboard camera was calibrated offline to get its internal matrix, it is ready for calculating pose data of unmanned aerial vehicle precisely.Then consider the important influence of landmark shape and size, Angular point geometry distribution and number of points on pose estimation precision, a "T" type landing landmark is designed with given size parameters, combine landmark contour extraction with corner detection algorithm to get eight corners with good geometric distribution and the number is reasonable for pose estimation to guarantee the posture calculation accuracy; Finally, Real-time flight pose parameters of UAV through the projection relationship between 3D space and 2D image plane are estimated.Secondly, considering real time and accuracy requirement the above algorithm is optimized. Pose estimation algorithm in the process of uav landing included image preprocessing, target tracking and pose computation. In order to guarantee real-time performance of algorithm, the Lucas- Kanade(LK) optical flow method tracking landmarks stably was made significant improvement. To reduce the processing time of Lucas – Kanade(LK) optical flow method tracking landmarks stably, using the extracted eight corners as LK optical flow method input to detect and track, ensuring Real-time performance of the algorithm. And we also use kalman filtering algorithm to improve the precision of pose estimation.Finally, the flight visual simulation experiment was designed, and the design process of the experiment, experiment framework, application process and the fly frame of simulation experiment were introduced in detail; And the results of simulation experiment is given; In the end Real-time performance of algorithm was analyzed.we can intuitive show the estimated and actual pose data in the process of unmanned aerial vehicle(uav) landing by combining visual navigation algorithm and visual simulation experiment,it recreates the landing process in multi-angle, multi-channel and verify the truthfulness, accuracy and feasibility of the algorithm. The results of simulation experiment show that the algorithm has high precision,and the average period is 55.5191 ms(about18.0 frames per second). The real-time requirements of visual aided navigation of autonomous landing at low speeds of landing stage is satisfied basically.