Synthetic aperture radar target identification method based on auxiliary decision update learning

Abstract

The invention belongs to the radar remote sensing application technology field and particularly relates to a synthetic aperture radar target identification method based on auxiliary decision update learning. According to the method, a small amount of initial training samples are utilized to train an initial model, newly-added unlabeled images are utilized as test samples, the identification results are taken as training samples for next training, on the basis of the existing model, iterative training is carried out till an identification system with stable and mature identification efficiencyis acquired. The method is advantaged in that the convolutional neural network is utilized as the main body to extract deep characteristics of SAR targets to carry out classification, in combination with auxiliary decision of an auxiliary classifier, so the newly-added unlabeled SAR images can be directly applied to the existing classifier, moreover, repeated sample training is avoided, and identification efficiency is improved.

Description

[0001] technical field [0002] The invention belongs to the technical field of radar remote sensing applications, and in particular relates to a synthetic aperture radar target recognition method based on auxiliary decision update learning. Background technique [0003] Synthetic Aperture Radar (hereinafter referred to as SAR) has the characteristics of all-time and all-weather, and is an important means of earth observation. SAR target recognition uses SAR image information to realize the determination of target types, models and other attributes. It has clear application requirements in military fields such as battlefield reconnaissance and precision strikes. It is one of the key technologies to improve the information perception ability of SAR sensors and realize the application of SAR technology. [0004] SAR target recognition performance is closely related to training samples. Target recognition requires a large number of samples with classification labels, which requi...

AI Technical Summary

Problems solved by technology

However, these studies are based on optical image data

However, the imaging mechanism of SAR images is very different from that of ordinary optical sensors, resulting in that SAR images cannot be intuitively understood like optical images. The newly added SAR images do not have classification labels, and it must be trained to confirm the radar images conveyed. information, and complete manual reading and understanding cannot meet the real-time requirements of some applications

At the same time, compared with optical images, the special imaging mechanism of SAR images causes certain distortions in SAR images, which makes feature extraction for SAR images difficult.

Images