Answers are grounded in this paper and cite their sources.
Recent advances in the defense and industrial sectors have increased the demand for situational awareness systems (SASs) that integrate AI and XR technologies. However, generating high-resolution 360° panoramic imagery with multi-camera systems remains challenging due to parallax-induced shape distortions of foreground objects. This study presents a content-preserving panoramic stitching framework tailored for SAS applications, focusing on minimizing perceptual object deformation and ensuring robust image quality in extreme environments. Unlike conventional feature-based methods, the proposed framework utilizes predefined structural priors of radially aligned multi-camera setups to enable fast and consistent alignment. Seam lines are dynamically updated in perceptually homogeneous regions using a cost function that incorporates edge density, spatial continuity, and temporal smoothness. In addition, a novel misalignment detection module is introduced to distinguish between parallax-related differences and true structural misalignments caused by disturbances such as vibration or shock, enabling real-time quality monitoring. The framework is experimentally validated on a custom-built scaled RC vehicle platform. Seam analysis in distortion-prone regions confirms that high-quality content-preserving panoramas can be produced faster and without relying on expensive semantic segmentation or object-level detection. Moreover, in synthetic perturbation scenarios, the misalignment detection module successfully excludes parallax-induced inconsistencies and accurately identifies only undesirable physical misalignments. These results support the effectiveness of the framework to maintain image quality in SAS and its applicability to real-world environments.
No takes yet. Share an insight, caveat, or question.
Cho et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: