空间机械臂对于非合作目标的视觉导航与跟踪研究

发布时间：2018-03-01 01:14

本文关键词： 非合作目标双目视觉目标识别位姿测量 Kalman滤波　出处：《北京理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：目前,以传统航天器的在轨维护、太空垃圾清理等为目的的空间非合作目标捕获技术成为空间机械臂领域新的发展方向。非合作目标的视觉导航与跟踪技术作为捕获过程中的关键技术更是各国研究人员的研究热点。为了解决非合作目标的视觉导航和跟踪问题,本文主要对非合作目标位姿测量和跟踪技术进行了研究。本文首先建立了基于双目立体视觉的位姿测量模型,然后研究了运动目标跟踪方法,最后通过实验仿真验证了以上模型和方法的有效性。主要内容如下:1)对二维图像进行处理,提出了基于特征信息融合的非合作目标识别方法。该方法针对非合作目标无法提供有效合作信息的问题,以航天器自有的特征信息作为识别对象,得到非合作目标矩形和圆形特征的基本形状信息,完成了目标识别和平面中心判定。并且根据尺度不变特征变换算法(scale invariant feature transform algorithm,sift)对左右相机获取的图像进行特征点匹配。2)研究了基于图像处理的双目视觉相对位姿测量方案。该方案选取目标卫星本体平面的中心点作为目标坐标系原点,建立非合作目标坐标系,计算相对位姿。然后建立了立体深度和视差的关系模型,分析了位姿计算的误差来源。最后选用OpenGL进行建模仿真,验证提出方案的可行性、精度与实时性。3)实现了非合作目标的视觉追踪。针对跟踪高机动运动目标的非线性问题,本文基于Kalman滤波算法引入IMM算法,加强了速度跟踪的性能。并且分别在目标匀速直线、匀加速直线和复杂曲线运动模型下比较了两种滤波算法的预测性能。具体实验分析证明,本文提出的非合作目标识别方法有效保证了目标识别的准确性,位姿测量方案和IMM滤波追踪方法准确、简便易行;算法的改进和模型的优化能较显著提高对非合作目标航天器的视觉导航跟踪精度。
[Abstract]:At present, with the on-orbit maintenance of traditional spacecraft, Space non-cooperative target capture technology, such as space garbage removal, has become a new development direction in the field of space manipulator. As a key technology in the acquisition process, the vision navigation and tracking technology of non-cooperative target is studied by many countries. In order to solve the problem of visual navigation and tracking of non-cooperative targets, In this paper, the non-cooperative target pose measurement and tracking techniques are studied. Firstly, the pose measurement model based on binocular stereo vision is established, and then the moving target tracking method is studied. Finally, the effectiveness of the above models and methods is verified by experimental simulation. The main contents are as follows: 1) the two-dimensional image is processed. A non-cooperative target recognition method based on feature information fusion is proposed, which aims at the problem that non-cooperative target can not provide effective cooperative information. The basic shape information of the rectangular and circular features of the non-cooperative target is obtained. Target recognition and plane center determination are completed. Based on scale-invariant feature transform algorithm scale invariant feature transform algorithm, feature point matching. 2) Binocular vision relative position based on image processing is studied. This scheme selects the center point of the body plane of the target satellite as the origin of the target coordinate system. The non-cooperative target coordinate system is established and the relative pose is calculated. Then the relationship model between stereo depth and parallax is established, and the error source of position and pose calculation is analyzed. Finally, OpenGL is selected for modeling and simulation, which verifies the feasibility of the proposed scheme. Aiming at the nonlinear problem of tracking high maneuvering moving targets, this paper introduces IMM algorithm based on Kalman filtering algorithm to enhance the performance of velocity tracking. The prediction performance of the two filtering algorithms is compared under the motion model of uniform acceleration line and complex curve. The experimental results show that the proposed non-cooperative target recognition method can effectively ensure the accuracy of target recognition. The position and attitude measurement scheme and IMM filter tracking method are accurate and simple, and the improvement of the algorithm and the optimization of the model can significantly improve the visual navigation tracking accuracy of the non-cooperative target spacecraft.
【学位授予单位】：北京理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V448.2;TP391.41

【参考文献】