The Research On Feature Extraction And Fusion Recognition For Radar Target

The early radar mainly used radio to find targets and acquire their location information.With the progress and development of the times,people want to explore more abundant information about radar targets,such as the type,attributes,structure,and so on.Therefore,the radar target recognition technology has emerged.It is a technology that extracts the stable and typical features of the target by analyzing the radar echo signal,and judges the type and attribute of the target by classifier.In recent years,with the rapid development of science and hardware equipment,the research of radar target recognition technology has attracted more attention from scientists all over the world.Nowadays,how to effectively improve the accuracy and reliability of radar target recognition is our main concern,and the research of radar target feature extraction and recognition methods are two key technologies.Therefore,this thesis mainly focuses on these two aspects.The main contents of this article are as follows:1.Aiming at the problem of extracting geometric dimension feature of radar target,the dimension estimation of target in high resolution Synthetic Aperture Radar(SAR)image is extracted based on the minimum enclosing rectangle method.On the basis of the SAR image denoising based on signal sparse representation theory,the reconstructed target image is obtained,and then the minimum enclosing rectangle of the reconstructed target is extracted.The dimension characteristic of the radar target is obtained from the size information of the minimum enclosing rectangle.Based on the measured data of vehicle targets,experiments show that the average relative error of target size feature extraction from high resolution SAR images is less than 5%,it is proved that this method has high accuracy.2.For space target recognition task,a metric learning feature fusion method is proposed combined with the common features of Radar target echo wave forms.Metric learning defines a similarity measure for data structure,and optimizes the similarity measure by using supervised information such as labels,so that the optimized metric space can achieve machine learning tasks such as clustering or classification.In this paper,the proposed method is applied to space target recognition.Firstly,the corresponding features are extracted from the high-resolution Range Profile(HRRP)and the Inverse Synthetic Aperture Radar(ISAR)images of space targets.Then,the fusion feature with optimal linear separability is obtained by metric learning.Finally,the fusion feature is input into the classifier for recognition.Based on the electromagnetic simulation data of space satellite targets,experiments show that the proposed method can significantly improve the separability of sample space,and has better target recognition performance than the method using only a single feature.