Space Target Recognition Based On Narrowband RCS Data

Space target recognition,that is to use radar to obtain echo signals of space targets,and then effective features are extracted from echo signals,and the type and attribute of space targets are identified by classifier.Detection and identification of space targets tend to be more and more essential for the development of space surveillance system,the use of space resources,and the guarantee of space security with the strategic position of space becoming increasingly prominent.This thesis focuses on the Radar Cross Section(RCS)data acquired by narrowband radar to conduct research on space target recognition technology.The main contents include:1.The definition of space targets' s RCS is introduced,and the scattering center model is mainly studied.Based on this model,the method of acquiring space targets' s RCS data is described,and the dataset is provided for subsequent space target recognition.Finally,the effects of attitude,angle and size on the RCS sequence of space targets are analyzed.2.A period estimation method of space targets' s RCS sequence by combine function and variance analysis is studied in this thesis to settle the problem of poor robustness and multiple frequency misjudgment in the process of space targets' s RCS sequence period estimation by time frequency analysis and auto correlation class functions.Firstly,the auto correlation function and the average magnitude difference function are improved and a combine function is constructed in this method.False peaks and valleys are suppressed effectively by the combine function and the anti-noise capability is good Compared with directly using auto correlation class functions.The multiple frequency misjudgment is reduced and the period feature of space targets' s RCS sequence is extracted effectively in this method when combined with analysis of variance.Simulation results show that the method studied in this thesis can improve the accuracy of period estimation.3.It is very difficult to directly derive the mapping relationship between the space targets' s RCS sequence and targets' s size,which results in high estimation error.A method of space targets' s RCS sequence size estimation based on deep learning is studied in this thesis.The mapping relationship between space targets' s RCS sequences and targets' s size is established automatically by deep learning in this method.Compared with traditional methods,the above mapping relationship is mined better and the size of the space targets is estimated effectively by deep learning method based on big data.Simulation results show that the method studied in this thesis can reduce the error of size estimation.4.The use of statistical features alone can not effectively identify space targets which will also lead to a problem of low recognition rate.Therefore,the space target recognition based on comprehensive features is studied in this thesis.In this method,combine function and variance analysis are used to extract periodic feature of space targets,and then deep learning is used to extract size feature of space targets.Compared with the statistical features,the performance is enhanced by periodic feature and size feature to represent the characteristics of the space targets.Finally,space targets are identified by the comprehensive features.Simulation results show that the method studied in this thesis can improve the accuracy and stable effect of space target recognition.