舰机相对位姿鲁棒、高精度测量是舰载机自主着舰引导的关键技术之一。基于关键点检测和Perspective-n-Points（PnP）求解的机载单目位姿测量方法因其配置简单、能耗低、无需数据链支持等优势受到研究人员的广泛关注，但已有方法未考虑关键点检测误差，难以实现鲁棒、高精度相对位姿测量。针对这一问题，将舰船目标建模为稀疏关键点集合，进一步转化相对位姿测量为飞机运动状态估计问题，将二维关键点检测结果作为观测量，构建紧耦合的扩展卡尔曼滤波系统；针对观测噪声统计特性未知的问题，提出基于滑动窗口的观测噪声协方差矩阵自适应估计方法，进一步提升算法精度。仿真实验和真实缩比实验结果表明：方法鲁棒性、位姿解算精度优于传统方法，实现了机载单目视觉着舰引导中鲁棒、高精度的在线相对位姿测量。

Robust and accurate relative pose measurement is one of the key technologies for shipboard aircraft autonomous landing. The airborne monocular pose estimation methods， based on key-points detection and Perspective-n-Points （PnP） problem solving， have drawn significant attention of researchers for its strengths in ease of deployment， power efficiency and anti-electromagnetic interference， etc. However， the inaccuracy of key-points detection is not considered in the existing methods， which corrupts the precision of the pose identification. To address this problem， a tightly-coupled AEKF-based monocular pose tracking method is proposed. The pose measurement problem is transferred into motion state estimation of the aircraft. An extended Kalman filter system is established taking the key-point detection results as observations， with the carrier represented in form of a sparse key-point set. For the unknown statistics of the observations， an adaptive noise covariance estimation method based on sliding window is put forward. Synthetic and real scaled experiments demonstrate that the proposed method achieves robust and accurate online pose tracking， superior to traditional approaches.