High-resolution PolSAR vehicle target recognition dataset

Polarimetric SAR (PolSAR) vehicle target data is an essential foundation for conducting research on the analysis and modeling of target polarization scattering characteristics, extraction of target polarization features, and automatic target recognition. For a long time, the scarcity of PolSAR measured data, low resolution, and insufficient diversity of scenes and target samples have severely limited the improvement of related research and application levels. In particular, with the rapid development of artificial intelligence technology, the scarcity of data samples has become a significant barrier to the intelligent recognition applications of PolSAR. To address this issue, we have constructed a full-polarization SAR vehicle target recognition dataset for the first time through a series of processes, including typical scene setup, UAV field flight, high-resolution imaging, and data cropping and preparation. This dataset covers seven major categories and 21 types of vehicles, with three typical backgrounds (concrete pavement, grassland, and shrubbery), two frequency bands (Ku and X), five depression angles, and up to 72 azimuth angles, comprising a total of 3,308 Pauli-RGB images and corresponding polarization scattering matrix data. It can provide high-quality data support for research on vehicle target polarization scattering characteristics, fine-grained recognition, few-shot learning, domain adaptation, and other studies, promoting the reliable application of PolSAR target recognition technology.