计算机视觉
人工智能
计算机科学
极化(电化学)
图像处理
遥感
光学
地理
物理
图像(数学)
化学
物理化学
作者
Aiqi Zhong,Qiang Fu,Danfei Huang,Jianxun Zhu,Huilin Jiang
标识
DOI:10.1080/10106049.2023.2204059
摘要
The performance of sole optical imaging device for marine target recognition is degraded by sea fog, sea-glint, and many other disturbing factors. To enhance the ability of target recognition in marine environment, we propose a polarization-intensity joint imaging method and the corresponding processing method. We combine fine imaging of polarization in a small field of view with wide-field imaging of visible light intensity, using visible light intensity information for large-scale target surveys. After locking onto areas of interest, we utilize high-resolution polarization cameras with small fields of view for detailed inspection, and enhance and fuse the information from areas of interest using the proposed matching information processing method. Meanwhile, to deal with problems existing in the marine environment polarization-intensity joint imaging, such as details loss in DoP (degree of polarization) image, low target resolution in the large field of view intensity image, and insufficient information in a single image, etc., we extract the details of raw image from DoFP (division-of-focal-plane) polarization camera as residual compensation for DoP image super-resolution and cooperate with the RealSR algorithm to super-resolve the local target details of the intensity image in a large field of view. On the premise that PIQE (Perception-based Image Quality Evaluator) reaches the excellent score range, the images with the same resolution are fused. The automatic recognition comparison before and after processing proves that the accuracy of target recognition can be effectively improved after processing.
科研通智能强力驱动
Strongly Powered by AbleSci AI