Researches On Monocular Vision Based Pose Measurements For Space Targets

With the increasing deepening of a new revolution in military affairs, the space has be-come the new commanding elevation of strategic competition, and the space security has be-comethecornerstoneofacountry’ssecurityanddevelopment. Thespaceoperationtechnologyisoneofthevital basic capacitiesforprotectingthespacesecurity, which playsanincreasinglyimportantroleinthespaceprogramandmission. Themeasurementoftherelativepositionandattitude parameters (pose parameters) from the target spacecraft to the tracking spacecraft isa key essential step of most space operation tasks. Monocular vision is one of the significantmethods to achieve the measurement process, which is the preferred means of measurementat close range. Therefore, it is necessary to do more in-depth research on space object posemeasurement technique based on monocular vision.The background of this paper is based on the obtaining of relative pose information of thetargetincloserangesegmentofthespaceoperationtasks. Itstudiesthemethodsofspaceobjectpose measurement based on the monocular vision system. The target of the study is to explorehow to improve the speed, accuracy, robustness and the adaptation capacity of space objectpose measurement. This paper avoids the problems such as image pre-processing and featureextraction which are closely related with the actual system and the target characteristics. Onthe contrary, it makes efforts to solve the problem of solving the object pose based on featurepoints on the algorithmic level. The main contents of the study are listed as follows:(1) It explores the Orthogonal iterative (OI) algorithm in depth, raises a improved real-ization form of the OI algorithm, and it proposes a new solution of the pose ambiguity causedby the planar targets. The improvement is realized by means of adjustment formula to reducecalculational redundancy, and of adopting paraperspective model instead of weak perspectiveduring the initial pose solution process. This method not only greatly improves the speed ofposition solution, but also further improves its stability and convergence performance. Thisstudyraisesanewsolutiontosolvetheposeambiguitycausedbytheplanartargets,whichisre-alized by the iterative optimization using the two local optimal pose under the paraperspectivemodel as the initial value. It works out a new resolvent, which is less an orthogonal iterativeoptimizationprocesscomparedwithexistingsimilaralgorithmandwhichgreatlyimprovestheefficiency of operations.（2）It studies the paraperspective (PI) iteration algorithm in-depth and it makes a lotof improvement on the PI algorithm. The performance of PI algorithm will be affected by the space distribution of the target state because it needs a special projection reference point. Thisstudy focuses on this problem and improves the convergence performance and accuracy of thePI algorithm through clever use of homogeneous coordinates to eliminate the dependence ofthe reference point. Although the PI algorithm provides a method to solve the pose ambiguity,it still takes a lot of time to choose the intermediate solutions in an iterative process. Theimproved algorithm proposed an intermediate solution fast selection strategy, and it solves anumerical problem in the original algorithm calculation process, which obviously improvesthe speed and stability to the pose solution of planar targets.（3）The study proposes a simultaneous pose estimation and correspondence determina-tion algorithm based on the paraperspective model (SoftPI). It works out a new algorithm byintroducing the paraperspective model into simultaneous pose estimation and correspondencedetermination solution, referencing the way of SoftPOSIT algorithm. The new algorithm hasa stronger local search ability and higher operation efficiency, on the premise of the similaraccuracy with the SoftPOSIT algorithm. Moreover, SoftPI algorithm could also solve the poseambiguity by applying the intermediate solution fast selection strategy raised in this paper. Itis the first time that an algorithm of the same kind has the ability to deal with a planar target.（4）This study proposes a simultaneous pose estimation and correspondence determi-nation algorithm based on differential evolution (DePose). SoftPI and SoftPOSIT have a poorglobal searching ability, which are more suitable for solving simultaneous pose estimation andcorrespondence determination problems with a known pose range. In the condition that thepose information is totally unknown, this study raises a better solution by introducing the dif-ferential evolution algorithm and proposing a new differential evolution mechanism and theover boundary parameter processing method. The research results indicates that the DePosealgorithm not only has higher computing efficiency and stronger performance of global searchthantheexistinggeneticalgorithmbasedsolution,butalsohashigherlocalsearchperformancethan SoftPI algorithm.It is aimed that the research findings of this study provides technical support to the spacemobile platforms for real-time accessing the short-range/close short-range morphological in-formation of the target in space. It is also quite easy to be promoted in other application fieldssuch as visual servo, mobile robot localization and navigation, and virtual reality.