Skip to content
Open access

Small-Sample Space Target Recognition Based on Polarimetric Scattering Mechanisms and Geometric Features

2026 · IEEE Transactions on Geoscience and Remote Sensing · Vol 64, pp. 5217914-5217914 · 0 citations · 49 references

Abstract

Space target recognition (STR) plays a crucial role in ensuring space security. However, existing methods exhibit insufficient exploitation of polarimetric information, limited robustness to angular variations, and heavy reliance on large-scale labeled data. To address these issues, inspired by the attributed scattering center (ASC), we propose a method that jointly exploits the physical scattering properties and geometric features for target recognition under limited training samples. To characterize scattering behavior, both the scattering mechanism and the diversity of polarimetric descriptions are important; therefore, polarimetric features are explored from two aspects. First, we design a collaborative attention mechanism that dynamically integrates polarimetric signatures with dominant scattering mechanisms to enhance the representation of discriminative scattering centers. Second, to alleviate cross-angle sensitivity, a subspace polarimetric descriptor is introduced to characterize the coherence statistics under different polarization states. Finally, the geometric contour features extracted by a shallow convolutional neural network (CNN) from binarized images are used to provide complementary structural priors for the physical scattering features. Experiments on the polarimetric space target dataset demonstrate that the proposed method achieves superior performance by effectively integrating scattering features and geometric priors within a lightweight architecture. Compared with existing methods, the proposed method achieves improvements of 10.91% and 16.3% in the 1-shot setting for the same-angle and cross-angle evaluations, respectively, and 1.24% and 3.87% in the 3-shot setting.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.