基于视觉的卫星在轨自维护操作的研究-机械电子工程专业论文.docx

了视觉伺服过程中可能出现的机器人奇异问题，提高了机器人操作的安全性。视 觉伺服算法融入了基于关节力矩的阻抗控制，使得当卫星自维护平台执行接触 操作时，视觉伺服能够提供与被操作物体的接触信息，从丽提高了卫星自维护平 台对工作环境的适应能力和顺从能力。最后，结合前文的视觉研究工作，为了解决空间机器人操作中通讯时延、通 讯带宽限制等问题，提出了一种基于地图的空间机器人操作方法。基于地图的空 间机器人操作方法采用了操作人员利用虚拟环境生成操作地图，机器人执行端 根据地图完成任务的方式。论文中对基于地图的空间机器人操作方法中涉及到 的虚拟环境建模，人机接口设计，地图生成，机器人如何依照地图进行操作等问 题展开了详细的论述。在本文的最后，利用实验验证了本文提出的基于地图的空间机器人操作方 法的正确性和有效性。关键词 模式识别; 视觉跟踪; 视觉伺服; 阻抗控制; 在轨自维护AbstractIf the satellite is able to maintained on-orbit or be serviced by itself, not only the life of the satellite can be extent, and its performance can be enhanced in space systems, but also the substantial costs of waste and the growing problem of space junk can be reduced. In recent years, the development of space robot technology shows that its great potential in the satellites self-servicing. The work of this dissertation is supported by Commission of Science Technology and Industry. All experiments are done on the satellite on-orbit self- maintaining system’s experimental platform developed by robotics institute, Harbin Institute of Technology. The major work focuses on pattern recognition, visual object tracking, visual servoing, and map-based operation for space robot.Image feature extraction plays an important role in pattern recognition, among which, Corner is a quite important feature. In our work, Harris corner detector is im- proved. On this basis, a corner-based feature selection method is proposed for the char- acteristics of the objects in satellite having obvious edge and corner as man-made ob- jects. New method reduce the required number of features. JointBoost has been widely adopted to detect different kinds of objects. It’s improved by combining multiple clas- si?ers in a cascade to get faster detection speed and higher classi?cation accuracy. The pose of the object should be calculated after the object is identi?ed so as to enable robot to complete the operation on the target object. To achieve higher precision, an optimized three-dimension coordinate on-line me