在航天器视觉相对导航中,为了能够跟踪目标航天器,并对其的位姿信息进行实时、精确地观测和估计,首先需要针对相关图像建立稳定而快速的特征点提取与匹配算法,而特征点提取和匹配算法的准确性与实时性直接影响了航天器相对位姿的估计精度。针对SIFT算法计算量大,匹配时间长,不能满足航天任务高实时性的问题,提出了B-Harris( Binary Harris)算法,该方法结合SIFT算法的结构思想,通过采用多尺度Harris算子提取具有尺度不变的特征点;同时采用128位二进制描述子建立特征向量,组成B-Harris算法,使得特征点匹配时间大大降低。最后通过与SIFT算法的实验对比,证明了B-Harris算法能够适应航天器发生的旋转、尺度、视角等变化,并且在实时性方面明显优于SIFT特征点算法,能够满足航天任务的实时性要求。
A method for designing robust tracking controllers is proposed in this paper for a class of large-scale in-terconnected linear systems with uncertainties and time-delays. If the uncertainties satisfy the matching conditions, we can construct a decentralized tracking controller by using the solution of the Riccati equation. With the decen-tralized controller, the closed-loop system will asymptotically track the reference input, even if the systems contain time-delays in both states and controls. Numerical simulation results and their analysis have demonstrated the effec-tiveness of the approach proposed.