Target detection enhancement method through multi-dimensional information differential projection and fusion

作者：Xue Dong, Yubo Lan, Zongmin Zhao, Chang Cai, Zhiming Zhou, Tong Zhang, Pingli Han, Meng Xiang, Jinpeng Liu, Peiyue Xie, Fei Liu · 发表于：Optics Express · 年份：2025 · DOI：10.1364/oe.560839 · 被引用次数：3 · 研究领域：Optical Polarization and Ellipsometry、Infrared Target Detection Methodologies、Remote-Sensing Image Classification

As a powerful detection tool, spectral polarization imaging (SPI) reveals spatial, spectral, and polarization characteristics. When combined with deep learning, its detection capabilities can be further improved. However, SPI requires the acquisition and processing of large-scale data, and it is still facing significant challenges in detecting targets in complex scenes with reduced contrast, loss of spatial structures, and cluttered backgrounds. Hence, an innovative target detection enhancement method via multi-dimensional information differential projection and fusion is presented based on our self-developed SPI systems. It primarily comprises a target retrieval algorithm based on angle of polarization purification and a novel projection and fusion strategy, which can effectively exploit, enhance, and utilize the multi-dimensional distinctions between the target and background. The resulting fused image visualizes these distinctions, thereby highlighting the target from its background. Experiment results demonstrate the efficacy of our method in detecting targets with shielding, camouflage and shape loss in complex natural environments, i.e., grasslands, woodlands, and sky. This significantly enhances scenario comprehension and detection efficiency without extensive computation or model training, making it especially advantageous for military reconnaissance applications.