Researches On Vision-Based Pose Measurements For Space Targets

In this dissertation, the methods and algorithms of vision based targets' pose measurements are studied and ground simulation experiments are conducted. The background of the works is to measure space targets' relative distance and attitudes for the two projects of space cabin docking and space robot application.Because of the harsh environment and lots of unknown factors in the space, videogrammetric system parameters tend to be disturbed. The disturbance will badly impact the results. To remove the infections of the disturbance, disturbance-rejecting methods of pose measurements are studied especially.The main contents of the works in this dissertation are as follows.1). Camera parameters' redundancy is analyzed. 7 kinds of precision-reserved simplified imaging models are concluded. By using these simplified models, lots of methods and algorithms in videogrammetry can be simplified without loss of precision. With these models, the requirement on control conditions can be reduced.2). Methods and algorithms of camera calibration are studied. Several new methods of calibration are proposed.Based on the precision-reserved simplified imaging models, a method of camera calibration by two iterated steps from coplanar control points and a method to calibrate a camera's linear parameters from coplanar control lines are proposed. Coplanar signatures are easier to be prepared than non-coplanar signatures. Thus the proposed two calibration methods are more convenient to be applied.For the calibration of cameras' exterior parameters relative to the basic coordinate system, a panorama method by using an assistant camera and a simulated docking method which is adapted to the docking application are preposed. For the two-step method of eye-hand calibration, an algorithm by resolving linear equations is proposed. This algorithm is intuitionistic and simple.3). The aberration of scan images by a common scanner is analyzed. Then a method to calibrate a scan image by distributing standard orthogonal griddings around the scan image is presented. This method makes normal scanners be used for accuracy measurement.4). Methods and algorithms, especially the disturbance-rejecting methods of targets' pose measurements are studied.A point filling method to estimate a cooperative target's pose linearly and roughly from 3 points is prposed. When the target's yaw angle and pitching angle are small, this method can provide initial values for adjustment.The following two kinds of disturbance-rejecting methods of pose measurements are proposed.One method is to fix the measuring camera with an assistant camera that observes the disturbace of the measuring camera's exterior parameters and corrects them in real time.The other method is to take the disturbed parameters as adjusted values and modify them by bundle adjustment. By using this method to measure cooperative or uncooperative targets' pose by monocular or binocular videogrammetric system, the disturbance on the cameras' interior parameters, aberration coefficients and relative pose parameters between the two cameras can be removed. These disturbance-rejecting functions are realized by arithmetic-software and do not require any change on the hardware system.By the above disturbance-rejecting methods, the problems fetched from the disturbance on the videogrammetric system parameters are solved. Thus accurate measurements can be reached even if the system parameters are disturbed.5). Ground experimental systems of vision-based pose measurements for space cabin docking and space robot systems are developed. Observe systems' schemes are designed and the data and physical simulation experiments are conducted.The methods and algorithms studied in this dissertation can also be applied on other videogrammetry tasks.