Department of Electrical & Computer Engineering, Biomedical Engineering, Applied Physics, Biointerfaces Institute, Macromolecular Science & Engineering, University of Michigan, Ann Arbor, MI 48109, USA.
Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI 48109, USA.
Sensors (Basel). 2022 Dec 28;23(1):323. doi: 10.3390/s23010323.
Dynamic detection in challenging lighting environments is essential for advancing intelligent robots and autonomous vehicles. Traditional vision systems are prone to severe lighting conditions in which rapid increases or decreases in contrast or saturation obscures objects, resulting in a loss of visibility. By incorporating intelligent optimization of polarization into vision systems using the iNC (integrated nanoscopic correction), we introduce an intelligent real-time fusion algorithm to address challenging and changing lighting conditions. Through real-time iterative feedback, we rapidly select polarizations, which is difficult to achieve with traditional methods. Fusion images were also dynamically reconstructed using pixel-based weights calculated in the intelligent polarization selection process. We showed that fused images by intelligent polarization selection reduced the mean-square error by two orders of magnitude to uncover subtle features of occluded objects. Our intelligent real-time fusion algorithm also achieved two orders of magnitude increase in time performance without compromising image quality. We expect intelligent fusion imaging photonics to play increasingly vital roles in the fields of next generation intelligent robots and autonomous vehicles.
动态检测在具有挑战性的光照环境中对于推进智能机器人和自动驾驶汽车至关重要。传统的视觉系统容易受到光照条件的影响，对比度或饱和度的快速增加或减少会使物体变得模糊，导致能见度降低。通过使用 iNC（集成纳米级校正）将偏振的智能优化纳入视觉系统，我们引入了一种智能实时融合算法来应对具有挑战性和不断变化的光照条件。通过实时迭代反馈，我们快速选择偏振，这是传统方法难以实现的。融合图像也使用在智能偏振选择过程中计算的基于像素的权重进行动态重建。我们表明，通过智能偏振选择的融合图像将均方误差降低了两个数量级，从而揭示了被遮挡物体的细微特征。我们的智能实时融合算法还实现了两个数量级的时间性能提升，而不会影响图像质量。我们预计智能融合成像光子学将在下一代智能机器人和自动驾驶汽车领域发挥越来越重要的作用。
