Recognition of Space Target Based on GNN under the Condition of Small Samples for ISAR Image
Haoxuan Yuan, Yun Zhang, Hongbo Li, Jiaying Chen, Muqun Niu · 2021 CIE International Conference on Radar (Radar) · 2021
With the development of space science and technology and the increasing demand for space monitoring, high-precision recognition technology of space targets has become more and more important. Inverse Synthetic Aperture Radar (ISAR) is an effective means of space detection, with the advantages of all-weather and high resolution. In recent years, there have been more and more techniques for applying deep learning methods to ISAR image recognition. However, for non-cooperative space targets, the lateral resolution of ISAR images is difficult to determine, and due to factors such as speckle noise and interference fringes, the quality of ISAR images will decrease to varying degrees. Therefore, the number of effective ISAR images obtained for the target is very small. This paper uses graph neural network (GNN) to find the deep relationship between similar samples by calculating the adjacency matrix between graph data nodes to realize small-sample recognition. Experimental results show that under the same conditions, the recognition rate of this method is about 10% higher than that of the CNN method with 20 times the sample size, which proves that this method can improve the reliability of spatial target recognition and has practical value.