基于深度自编码高斯混合模型的航天器有效载荷遥测数据融合方法

Spacecraft payload telemetry data fusion method based on deep autoencoder Gaussian mixture model

摘要:

航天器有效载荷遥测数据受太空环境影响而复杂多变,高纬度下包含的特征量众多,不能准确识别遥测数据中的有效特征,导致划分到不同类型中的数据集无法准确反映原始数据的特征和关系,影响融合质量。为此,提出基于深度自编码高斯混合模型的航天器有效载荷遥测数据融合方法。按照时间顺序对多站遥测数据进行全帧编号,并通过编号对齐实现数据采集时间统一；引入主成分分析特征选择算法,对有效载荷遥测数据中的关键特征进行有效筛选；构建深度自编码高斯混合模型,对航天器有效载荷遥测数据进行分类；基于支持因子的证据理论融合算法分析样本相关性,实现遥测数据的融合处理。实验结果表明:通过该方法完成航天器有效载荷遥测融合处理后,完整涵盖了不同测站获取的对地观测卫星遥测数据包含细节,融合数据与真实数据之间的相关系数超过了0.98,极大提高了数据处理质量。

Abstract:

The telemetry data of spacecraft payloads is complex and variable due to the influence of the space environment. At high latitudes, there are numerous feature quantities contained, which cannot accurately identify the effective features in the telemetry data. This results in datasets divided into different types that cannot accurately reflect the features and relationships of the original data, affecting the quality of fusion. Therefore, a spacecraft payload telemetry data fusion method based on deep autoencoder Gaussian mixture model is proposed. Number multiple telemetry data frames in chronological order and achieve unified data collection time through numbering alignment; Introducing principal component analysis feature selection algorithm to effectively screen key features in payload telemetry data; Construct a deep autoencoder Gaussian mixture model to classify telemetry data of spacecraft payloads; The evidence theory fusion algorithm based on support factors is used to analyze sample correlation and achieve fusion processing of telemetry data. The experimental results show that after completing the telemetry fusion processing of spacecraft payloads using this method, the telemetry data of ground observation satellites obtained from different stations, including details, are fully covered. The correlation coefficient between the fused data and the real data exceeds 0.98, greatly improving the quality of data processing.