Target Feature Extraction And Recognition In SAR Images

Synthetic Aperture Radar(SAR)is used mainly for the two-dimensional high-resolution imaging of the observed scene.Due to its application in various fields such as aeronautics and astronautics,SAR image interpretation technology has become a hot research topic.Target feature extraction and recognition are among the important research directions.SAR image target feature extraction is mainly to obtain the target geometric features and attribute features from SAR images,which can be utilized to the classification and recognition of targets.This paper focuses on SAR image target feature extraction study,including height estimation of the ground target,geometric feature extraction of ship target and SAR target attribute learning method,and achieves the software implementation of related algorithms.The main contents of this paper are summarized as follows:1.The height estimation of the ground target is introduced.Based on the radar imaging principle,the geometric relationship between the height of the target and the shadow area is found by analyzing the relationship between the 3-D real scene and the corresponding 2-D SAR image,so as to estimate the height of the target.Some prior knowledge is used to judge the reliability of the results of the feature extraction and to modify the results,and the height estimation result of the target is achieved.2.Geometric feature extraction of ship target is studied in this part.First,the principal component analysis(PCA)is performed on the target area to extract the principal axis of the target in order to obtain the length and width characteristics of the target using the traditional ship size estimation algorithm.Second,the traditional ship size estimation algorithm is improved,and the Convolutional Neural Network(CNN)and the Gradient Boosting Decision Tree(GBDT)are applied to the ship size estimation method.The model of this paper is used to estimate the size of the ship target,and the size is corrected based on the multi-dimensional scattering characteristics.Using the measured data,the two algorithms are compared experimentally.3.SAR target attribute learning method is studied.First,traditional attribute learning method is introduced.The artificially defined SAR target attribute features are learned through the SVM classifier to obtain the target attribute feature vector,which is used in target classification and recognition.Then,the SAR target attribute learning method based on CNN is introduced,and the results are applied to target classification and recognition.The results of the two methods are compared experimentally.4.Software of SAR image target feature extraction is designed.Firstly,the overall architecture of the software is introduced,including the software development architecture,the module functions of the software,and so on,and each function of the software is introduced in detail.Then,the design and implementation of the software data processing flow are introduced,and the software installation method is introduced.