Static and Dynamic Hyperspectral Image Analysis and Synthesis

静态和动态高光谱图像分析与合成

• 批准号: RGPIN-2019-04273
• 负责人: Yang, Herbert
• 金额: $ 3.5万
• 依托单位: University of Alberta
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=689732
• 关键词: Static; Dynamic; Hyperspectral; Image; Analysis

My long term research goal is to explore opportunities to create novel algorithms using physically correct methods for computer vision (CV) and for modelling and animating physical phenomena. Physically correct methods refer to computer vision methods that are consistent with principles in optics and in geometry. My underwater imaging work motivates my proposed research into investigating the computer vision issues of refraction of more than 3 wavelengths. The refraction of light occurs when light traverses from one medium to another medium with a different refractive index. The value of refractive index depends on the wavelength of light and on the temperature of the medium. Cameras are designed based on the principle of refraction. Refraction also occurs naturally in transparent objects such as in glasses, and in water. Mirage is a naturally occurring phenomenon in which the temperature gradient of air causes the refraction of light, which is also known as thermal-optical or photothermal deflection. Interestingly, by clever manipulation of the thermal gradient, an object can be made invisible. The area of CV that uses multiple wavelengths or bands is called multispectral imaging, when the number of bands N is small (N****For objective 1, issues with MH images will be identified and existing constraints used in CV will be evaluated for application in MH imaging. Specifically, I expect that there will be issues related to the image formation process, e.g. refraction, or to the type of camera. For objective 2, I plan to select several important existing CV algorithms, in particular, those that were developed in my lab, and extend them to the MH environment. For objective 3, I plan to develop structured light systems that use more than 3 wavelengths. For objective 4, I plan to investigate MH features captured in a dispersive environment that are useful for classification and other computer vision algorithms, e.g. shape from shading.**

我的长期研究目标是探索使用物理正确的方法创建新算法的机会，用于计算机视觉（CV）以及建模和动画物理现象。物理上正确的方法是指计算机视觉方法与光学和几何学原理一致。我的水下成像工作激发了我对超过3个波长折射的计算机视觉问题的研究。当光从一种介质穿越到另一种具有不同折射率的介质时，就会发生光的折射。折射率的值取决于光的波长和介质的温度。相机是根据折射原理设计的。折射也自然发生在透明物体中，如玻璃和水中。海市蜃楼是一种自然发生的现象，其中空气的温度梯度导致光的折射，这也被称为热光学或光热偏转。有趣的是，通过巧妙地操纵热梯度，物体可以变得不可见。使用多个波长或波段的CV区域称为多光谱成像，当波段数N较小时（N*）。对于目标1，将识别MH图像的问题，并评估CV中使用的现有约束，以应用于MH成像。具体来说，我预计会有与图像形成过程相关的问题，例如折射或相机类型。对于目标2，我计划选择几个重要的现有CV算法，特别是那些在我的实验室开发的算法，并将它们扩展到MH环境。对于目标3，我计划开发使用3个以上波长的结构光系统。对于目标4，我计划研究在分散环境中捕获的MH特征，这些特征对分类和其他计算机视觉算法有用，例如从阴影中提取形状。