Year: 2026
Type: article
Abstract: The recognition of SAR vehicle targets holds significant practical value in battlefield situational awareness and precision targeting. To address the scarcity of real SAR vehicle target data, the homogeneity of backgrounds, and limited sample diversity, we constructed a dataset of high-resolution SAR vehicle target recognition. This dataset was developed through the deployment of typical scenarios, field UAV flight tests, SAR data imaging processing, and preparation of SAR target slices. It comprises 7 categories and 21 types of vehicle targets, featuring three background types—cement ground, grassland, and shrubbery. Data are provide in two frequency bands, Ku and X, with five depression angles and up to 72 azimuth angles, resulting in a total of 3,308 SAR amplitude images and corresponding complex data. All samples are subjected to a dual verification process involving expert interpretation and image quality assessment to ensure the reliability of imaging quality and annotation information. This dataset provides foundational data support for research on SAR vehicle target characteristics analysis, fine-grained vehicle recognition, few-shot recognition, and domain-adaptive target recognition, thereby advancing the research level and application in the SAR target recognition field.
Source: China Scientific Data
Authors: Wei WANG, Huiqiang ZHANG, Hongqi FAN, Zhongjie WU, Bingwei HUI, Tianpeng LIU, Shujie WU, Deliang XIANG, Chaofu WEI, Yingbing MA
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7125209196