Research And Implementation Of Target Recognition Algorithm For Step-frequency Radar Imaging

Radar target recognition technology is an indispensable link in weapon system operations in the context of modern warfare.Its workflow is to process the echo signals received by the radar,extract relatively stable features that can reflect the distinguishing characteristics of the target,and determine it by the classifier Target category.For occasions where the platform is limited and high real-time performance is required,the use of stepped frequency radar with good performance,strong flexibility and low cost for high-resolution imaging is one of the important technologies to achieve target recognition.This thesis mainly analyzes the feature extraction and recognition performance of the imaging information obtained by the stepped frequency radar echo,and improves the target recognition algorithm based on feature fusion and deep learning technology to achieve the effect of improving the recognition performance.Combined with Huarui-2 DSP chip,a radar sub-system that uses High Resolution Range Profile(HRRP)and Inverse Synthetic Aperture Radar(ISAR)images for target classification and recognition under the stepped frequency system is designed.The research content of this thesis is as follows:1.The principle introduction and simulation verification of radar one-dimensional HRRP and two-dimensional ISAR imaging algorithms under stepped frequency system.According to the principle of waveform design,FEKO electromagnetic simulation software is used to simulate four types of aircraft target echoes.After processing by imaging algorithms,a large number of one-dimensional HRRP and two-dimensional ISAR images are obtained to form a data set for subsequent feature extraction and target recognition in this article.2.Combined with the data set in this article,the typical features of various target images are extracted,and the K-Nearest Neighbor(KNN)classifier is used to verify the recognition effect of a single feature.According to the number of training data samples,it is divided into cooperative and non-cooperative targets,and experimental analysis of feature recognition performance and the compressibility of the number of training samples are carried out respectively.3.Aiming at the single feature recognition defect,a metric learning algorithm based on Difference of Convex Functions(DC)programming is given,and the performance of the extracted features is improved and fused.It also uses deep learning technology to automatically learn image features to improve the performance of two-dimensional image feature recognition.4.Based on the Huarui-2 DSP,a radar sub-system using HRRP and ISAR images for target classification and recognition under the stepped frequency system is designed.The software and hardware platform used by the radar system is introduced,the working scheme of the signal processor of the sub-system and the DSP realization process are designed,and the various functions of the sub-system are verified by simulation software.