Research On Location Algorithm Of Asteroid Lander Based On Visual SLAM

Asteroid exploration is an important part of modern human exploration of the universe,in which asteroid lander positioning technology is an important branch of modern space probe navigation status research.Among the many navigation and positioning methods,the navigation and positioning method based on Visual Simultaneous Localization and Mapping(SLAM)combines the navigation methods in various fields,such as classical geometry,image processing technology,computer science and so on.However,due to the lack of scale information,the monocular vision SLAM algorithm is unable to estimate the distance between the features in the image.The combination of vision and inertial measurement unit is a suitable solution,compared with the binocular SLAM,the former is not limited by the baseline length,and the displacement is more accurate.In the special outer space environment,the asteroid lander cannot rely on its own sensors to locate actively with the help of global positioning system,base station positioning and other positioning methods.In this paper,around the asteroid lander landing section positioning algorithm,considering the use conditions and combination of sensors,the experimental equipment of visual inertial integrated navigation system based on filtering method is made,and the robustness and positioning accuracy of the system are improved.The main research content of this paper includes the following three aspects:First of all,aiming at the existing positioning methods in the landing section of asteroid lander,such as computational complexity,positioning feasibility and positioning accuracy,combined with the advantages of visual SLAM and Inertial Measurement Unit(IMU),the inertial information is fused into the visual SLAM system.To achieve visual inertial combination positioning,give full play to the respective advantages of vision and IMU,IMU can give scale information for the image,does not systematically provide continuous recurrence information,the image can eliminate IMU drift error.In the back end of visual SLAM,the final state estimation is obtained by fusion of visual and inertial information by Kalman filter.Secondly,in view of the fact that the visual SLAM is easily affected by the grayscale change before and after image special diagnosis,Self-adjusting Desensitized Filter(SDF)is described,and the sensitivity information is added to the system cost function.The sensitivity is adjusted in the algorithm for calculating the filter gain.At the same time,according to the filter convergence standard,the selfbalancing parameters are established to adjust the sensitivity and adjust the specific gravity.The desensitization filter with self-balance adjustment has better robustness and will not be too dull because of over adjustment.Finally,in order to solve the problem that the IMU deviation in the visual inertial SLAM system is not constant and the estimation is not accurate,Nonlinear Unbiased Filter(NUF)is proposed to construct Unbiased Visual SLAM(UVI SLAM)to estimate the IMU deviation in real time to reduce the influence of the estimation deviation on the estimation accuracy.At the same time,based on the unscented Kalman filter to deal with the linearization process,avoid the loss of accuracy,and finally improve the accuracy of the visual inertial combination positioning algorithm.