Image-Based Small Body Parameters Estimation And Spacecraft Autonomous Navigation

Exploration of the unknown space is the pursuit of development and progress of human society. Small bodies (such as asteroids and comets) exploration becomes the research focus of new century deep space exploration. The images taken in the exploration mission can provide small body shape and terrain surface directly as well as the navigation information of the spacecraft. How to make use of the image information to get more science and navigation information becomes one of the most important key technologies. With the supports of 863 Program (Asteroid Attachment and Experimental Demonstration System) and National Natural Science Foundation of China (Theory and Method of Deep Space Autonomous Navigation), this dissertation deeply studies the small body parameters estimation and spacecraft autonomous navigation based on the image information. The main contents of this dissertation are as follows:First, for the noisy images acquired from the cruise and approach phases, this paper gives radiometric calibration, geometric calibration and scattered light calibration to eliminate the noise. The bilateral filter rather than classical median filter is adopted to eliminate the Gaussian noise as well as to preserve the edges information of stars. Minimal circumscribed ellipse algorithm is presented to estimate the centroid of the small body. Furthermore, the target body tracking and small body rotation period estimation algorithms are presented.Second, this paper studies 3D model reconstruction of the small body and the physical parameters estimation method based on fly-around images and one distance information. In order to avoid the inefficient manual feature matching and reduce the effect such as light and rotation, this paper uses PCA-SIFT algorithm to match the feature points. Camera model can be simplified by weak perspective projection as the images are taken by narrow angle camera. The 3D small body shape and the state of spacecraft can be estimated by using SVD factorization method at the same time. After getting the position of the feature points, this paper presents a simple rapid surface triangulations algorithm. Base on this, the volume integral can be transformed into polyhedron vertex computation, so that the physical properties such as volume, surface area, center of mass, moment of inertia, principal axis of inertia, gravitational potential and gravitational field intensity of the small body can be computed.Third, this paper presents an autonomous landing navigation algorithm taking the scale invariant feature points as the landmarks. At the beginning of landing, the position and attitude of spacecraft can be estimated by using feature matching with the 3D shape features database builded during the rotation phase. At the terminal landing, the mosaic is taken as the virtual map, and the 6DOF state of spacecraft can be estimated even if the landing site can not been seen. On the other hand, by fusing feature matching and inertial navigation, the disadvantage of the feature matching algorithm can be overcome, and the cumulative error of inertial navigation can be reduced too, so that the landing reliability is improved.Next, in order to improve the stability and accuracy of the autonomous navigation, an algorithm of rapid state estimation between adjacent frames is presented. A multi-scale feature extraction algorithm, which computes the maximum of the features in moving windows, is used to improve the real-time performance. Simultaneously, this paper introduces an adaptive sampling algorithm which can compute the size and distribution of sample areas usefully to reduce the redundancy of computation. Base on these, the state equation is established in the essential space and the state of spacecraft can be estimated by using IEKF with the restriction of epipolar geometry.Finally, this paper presents terrain surface recognition, hazard areas extraction and autonomous safe landing site decision algorithms. The shadow and bright areas can be segmented by Top-Hat and Bottom-Hat transform. The terrain of the other areas can be reconstructed with recursive multiple view geometry and the orthography factorization algorithm. This paper also analyzes the application range and reconstruction precision of different methods. After ascertaining the hazard areas, the genetic algorithm and morphology algorithm are introduced to decide the safe landing site.