Synthetic Aperture Radar (SAR) is an active microwave imaging radar to obtain target image. Due to the character of robust performance under all-weather conditions and features of reflection of radio waves from various surfaces, SAR is widely used in military and civilian fields. In contrast to optical images, due to the physical principle of operation of the radar, the multiplicative noise is present in the radar images. This noise, also called speckle, significantly complicates the classification of objects in the image. The object orientation normalization is applied in addition to filtration using traditional classifiers such as SVM. Comparative analysis of traditional classifiers with different filters and methods of normalizing the orientation of the object has been formed previously. The attempt to use neural networks in various areas, where traditional methods dominated, it is a high research trend now. Because of this, various architectures of neural networks have been proposed to enhance the efficiency on synthetic aperture radar automatic target recognition (SAR-ATR) and obtained state-of-art results on targets classification in many articles. Most of the results were obtained using a widely used Moving and Stationary Target Acquisition and Recognition (MSTAR) database. Therefore, MSTAR database is used in this paper for easy comparison of the results obtained. In this paper, we compared the effects of various methods for preprocessing radar images for traditional methods such as SVM, decision trees and methods based on neural networks.
The present work is devoted to comparing the accuracy of the known qualification algorithms in the task of recognizing local objects on radar images for various image preprocessing methods. Preprocessing involves speckle noise filtering and normalization of the object orientation in the image by the method of image moments and by a method based on the Hough transform. In comparison, the following classification algorithms are used: Decision tree; Support vector machine, AdaBoost, Random forest. The principal component analysis is used to reduce the dimension. The research is carried out on the objects from the base of radar images MSTAR. The paper presents the results of the conducted studies.