自适应航天器态势分析系统

利用深度学习中的卷积神经网络理论,基于单目视觉系统和带有标识物的航天器影像,实现对航天器的三维姿态角、距拍摄点距离和相对拍摄中心偏移量的精准测量。利用机器学习理论实现网络自主学习样本特征,这一方式将大幅降低动态测量的误差。同时,这种测量方式也避免了人工提取特征的复杂过程,实现任意、精准、快速测量,对航天器在组装及发射过程中的姿态估计、距离测算起到关键性作用。

Adaptive Spacecraft Situation Analysis System

Based on the monocular visual system and the spacecraft image with the marker, this work uses the Convolutional Neural Network(CNN) theory in deep learning to accurately measure the three-dimensional attitude angle of the spacecraft, the distance from the shooting point and the offset from the shooting center. We take the advantage of machine learning theory to realize the network autonomously learning characteristics of samples, which will greatly reduce the error of dynamic measurement. At the same time, this method also avoids the complicated process of manually extracting features, realizing arbitrary, accurate and rapid measurement, and plays a key role in the attitude estimation and distance measurement of the spacecraft during assembly and launch.