一种卫星对地观测需求分析方法

A Method of Earth Observation Requirements Analysis for Satellite

针对卫星对地观测任务需求分析方法研究的复杂性和迫切需要,文章通过把文本化的卫星观测需求编码成数值向量的形式,提出需求之间的相似度度量方法。进而基于无监督学习的相关理论和算法,给出向量化表示的需求聚类算法。同时为了能在二维平面直观展示分析高维需求数据,在引入需求之间距离计算方法的基础上介绍了高维数据的降维算法,并且通过测试数据验证了算法的有效性和可行性。通过把繁杂的需求信息转换成高维空间向量,能大大降低分析人员对领域专业知识的依赖,提升需求分析的效率和科学性。

In view of the complexity and urgent need of requirements analysis of satellite earth observation tasks,this paper proposes a similarity measurement method by encoding the textual satellite observation requirements into the form of numerical vectors.Then,based on relevant theories and algorithms of unsupervised learning,a requirements clustering algorithm is given.In addition,in order to visualize and analyze the high-dimensional requirements data on the two-dimensional plane,the dimensionality reduction algorithm of the high-dimensional data is introduced in detail on the basis of introducing the distance calculation method between the requirements.The effectiveness and feasibility of the algorithm are verified using test data.By converting complicated requirements information into high-dimensional space vectors,it can greatly reduce analysts'dependency on domain expertise and improve the efficiency and scientificity of requirements analysis.