Target Detection Based on Regional Feature Difference in Synthetic Aperture Interferometric Radiometer

Bo Fang, Dong Zhu, Fei Hu, Yanyu Xu, Yakai Hao, Xinpeng Chen, Jingyu Tao, Jiale Min, Bolun Zheng

Abstract: The target detection by synthetic aperture interferometric radiometer (SAIR) has attracted attention, given its ability to provide instantaneous observation under poor visible conditions. Existing detection methods relying on a single type of feature suffer from false alarm pixels or missed targets. To address these issues, this article proposes a local region contrast-based (LRC) method using regional differences in the features of multiorder information. First, we analyze the dissimilarities in the zero-order, first-order, and second-order information using the Fisher vectors (FVs) in the differential brightness temperature (BT) images. Subsequently, we construct an FV contrast to depict target regions and avoid isolated false alarm pixels. Based on the dissimilarity characteristics of the aforementioned information features, a local region contrast feature is formulated for the target detection. Simulation and real-measured experiments validate the effectiveness of the proposed LRC method in SAIR target detection tasks.