Small sample SAR image target identification method

Abstract

The invention discloses a small sample SAR (synthetic aperture radar) image target recognition method, which comprises the following steps of: firstly, establishing a generative adversarial network based on the characteristics of an SAR image target, then introducing two types of objective parameters, evaluating a generated image, and screening an image with relatively high evaluation to construct an expansion data set; and finally, performing target identification on the expanded data set by using a CNN-based SAR image classification model, and comparing and analyzing results of the data sets expanded by various different methods. According to the method, the data distribution uniformity and content diversity of the SAR target recognition data set can be effectively improved, and compared with a traditional SAR data set capacity expansion method such as image cutting, scale transformation and inversion transformation, the method has the advantage that great improvement is achieved.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and in particular relates to a SAR image target recognition method. Background technique [0002] Synthetic aperture radar (SAR) has been widely used in various aspects of military and civilian fields due to its strong penetrating power and weather adaptability. hotspots. With the substantial improvement of SAR sensors, the requirements for SAR data processing are increasing day by day, and more and more SAR automatic recognition technology (Automatic Target Recognition, ATR) is put into use or enters the preparation stage. Coinciding with the popularity of deep learning in recent years, a large number of new models and new network architectures have been proposed, which can be used in ATR technology. However, the model construction of the existing target recognition algorithms often requires a large amount of training data to ensure the reliability of the model, and due to the partic...

AI Technical Summary

Problems solved by technology

However, the model construction of the existing target recognition algorithms often requires a large amount of training data to ensure the reliability of the model, and due to the particularity of the SAR image imaging principle, there is not enough expert-labeled data sets for the training model , which means that training the model with a small number of samples will bury hidden dangers such as overfitting, which will affect the final practical model recognition accuracy

Images