CRII: RI: Computational Thermal Imaging

想象一下，在未来，自动驾驶汽车可以穿透浓雾，甚至可以看到拐角处。有了这种超人的感知，可以避免多少事故，挽救多少生命？今天，普通的计算机视觉系统不能推广到恶劣天气或非视线障碍物规避。这是因为普通的计算机视觉系统依赖于可见光相机，这些相机可以看到我们眼睛看到的颜色。为了克服这一障碍，调查小组将研究计算热成像，其中使用红外光学和视觉算法设计新颖的端到端系统。这些问题通常是分开研究的。该研究计划将被整合到加州大学洛杉矶分校的一系列新设计的计算机视觉和计算成像课程中。该项目将计算成像带入长波红外波长的领域，并研究其独特的性能权衡。这项研究基于三个支柱。在第一个支柱中，研究团队将构建一个3D相机原型，该原型将在长波红外波长（约10微米波长的光）下投射定制设计的结构光图案。结构化光图案可以专用于长波红外波长的独特特性。在第二个支柱方面，调查小组将更深入地探讨在长波红外波长下工作的非视线照相机的可能性。创建一个可以看到角落的相机听起来像是一个艰巨的挑战，但调查小组的目标是利用长波红外波长特有的反射镜面反射。最后，研究小组将研究科学家们在长波红外波长和可见光波长下如何数学表示图像数据之间的根本差异。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Imagine a future where autonomous vehicles can see through thick fog or even around corners. How many accidents could be avoided, how many lives could be saved, with this form of superhuman sensing? Today, ordinary computer vision systems do not generalize to bad weather or non-line-of-sight obstacle avoidance. This is because ordinary computer vision systems rely on visible light cameras that see colors that our eyes see. To overcome this roadblock, the investigative team will study computational thermal imaging, where novel end-to-end systems are designed with infrared optics and vision algorithms. These topics are usually studied separately. This research program will be integrated into a sequence of newly designed courses in computer vision and computational imaging at the University of California, Los Angeles.This project brings computational imaging to the realm of long-wave infrared wavelengths and studies the unique performance tradeoffs thereof. The research rests upon three pillars. In the first pillar, the investigative team will construct a 3D camera prototype that projects custom-designed structured light patterns at long-wave infrared wavelengths (circa 10 micrometer wavelengths of light). The structured light patterns may be specialized to the unique characteristics of long-wave infrared wavelengths. In the second pillar, the investigative team will probe deeper into the possibility of non-line-of-sight cameras that operate at long-wave infrared wavelengths. Creating a camera that can see around corners sounds like a daunting challenge, but the investigative team aims to exploit the specularity of reflections that are unique to long-wave infrared wavelengths. Finally, the investigative team will study the foundational differences between how scientists mathematically represent image data at long-wave infrared wavelengths as compared to visible light wavelengths.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.