Self-learning method for SAR (Synthetic Aperture Radar) target recognition with incomplete visual angle of training target

Abstract

The invention discloses a self-learning method for SAR (Synthetic Aperture Radar) target recognition with incomplete visual angles of a training target. The method comprises the following steps: collecting a training set; two SAR images are extracted from the training set; taking the two SAR images as input information, determining a classification error based on the target recognition network, and determining a missing prediction error and a puzzle reconstruction error based on the missing prediction network and the puzzle reconstruction network; optimizing the target recognition network, the missing prediction network and the puzzle reconstruction network through the classification error, the missing prediction error and the puzzle reconstruction error; according to the method, the SAR images of different categories are randomly combined to generate the input pair, the relation between the visual angle incomplete SAR image and the visual angle complete SAR image is established, the diversity of the sample SAR images is greatly enriched, the target recognition network is optimized in combination with the missing prediction network and the jigsaw reconstruction network, and the target recognition efficiency is improved. Therefore, the network can extract robust identity features under the condition that the visual angle is incomplete, and the classification accuracy of a classifier obtained through training is improved.

Description

technical field [0001] The invention belongs to the technical field of synthetic aperture radar target recognition, in particular to a self-learning method for SAR target recognition with incomplete viewing angles of training targets. Background technique [0002] Synthetic Aperture Radar (SAR) target recognition is of great significance in military target recognition, vehicle re-identification and other tasks, and can provide accurate target identity information for monitoring, strike, rescue and other tasks. The most important task of target recognition is to provide stable and accurate identity information, and the key technology is the stable target recognition technology under the condition of incomplete perspective. In practical applications, synthetic aperture radar has the advantages of not being disturbed by conditions such as weather and light, and can continuously monitor targets in the environment. [0003] In actual target monitoring, the target and the satelli...

AI Technical Summary

Problems solved by technology

[0003] In actual target monitoring, the target and the satellite carrying the imaging radar are in relative motion. When the target is stationary or moving in a small range, it is impossible to obtain the imaging information of the target to be identified under all viewing angles, and only part of the viewing angle can be obtained. Imaging information, and SAR images are sensitive to viewing angles, small changes in observation angles can also cause sudden changes in SAR images

In practical applications, it is possible to obtain imaging of a part of the target at all angles of view (such as one's own target, etc.), but facing an imaging target with only a part of the angle of view, how to use the SAR image with imaging information of all angles of view and the imaging information of only a part of the angle of view? SAR images train a stable and accurate target recognition classifier, so as to deal with the problem of target recognition under the condition of incomplete viewing angle, which has become an urgent problem to be solved.

Images