Hyperspectral characterisation of natural illumination in woodland and forest environments
Hyperspectral characterisation of natural illumination in woodland and forest environments
复制标题
DOI：
10.1117/12.2595301
复制
发表时间：
2021-07
期刊：
影响因子：
--
通讯作者：
Liu Shiwen;Laura Steel;C. Dahlsjö;S. Peirson;A. Shenkin;Takuma Morimoto;H. Smithson;M. Spitschan
中科院分区：
文献类型：
--
作者：
Liu Shiwen;Laura Steel;C. Dahlsjö;S. Peirson;A. Shenkin;Takuma Morimoto;H. Smithson;M. Spitschan
自然界中的光是复杂的、动态的，它随光谱、空间、方向和时间而变化。虽然光谱分辨测量和空间分辨测量都是广泛可用的，但光谱分辨和空间分辨测量在技术上更具挑战性。在这里，我们提出了一个便携式成像系统，使用现成的组件，以光谱和空间分辨率的方式捕捉全球面光环境。该方法利用商业高光谱相机（400-1000 nm）从多个方向对镜面铬球反射的4π立体面光场进行成像，并通过图像处理流水线对镜面球进行提取、去除饱和像素、校正镜面球的镜面反射率、提升到高动态范围、校正图像的不对准、校正强度压缩、对成像系统进行擦除。展开球形图像，填充空白区域，拼接从不同角度采集的图像。我们将我们的方法应用于威瑟姆森林，这是英国牛津附近的一个古老的半自然林地。我们在两个低灰丰度和高灰丰度的地点获得了168幅图像，导致冠层的差异，总共获得了14个高光谱探测器。我们的图像处理管道充分纠正了小的（<3°）基于场的偏差。我们的新型高光谱成像方法适用于现场条件，为捕捉光环境的复杂性和动态性开辟了新的机会。
Light in nature is complex and dynamic, and varies along spectrum, space, direction, and time. While both spectrally resolved measurements and spatially resolved measurements are widely available, spectrally and spatially resolved measurements are technologically more challenging. Here, we present a portable imaging system using off-the-shelf components to capture the full spherical light environment in a spectrally and spatially resolved fashion. The method relies on imaging the 4π-steradian light field reflected from a mirrored chrome sphere using a commercial hyperspectral camera (400-1000 nm) from multiple directions and an image-processing pipeline for extraction of the mirror sphere, removal of saturated pixels, correction of specular reflectance of the sphere, promotion to a high dynamic range, correction of misalignment of images, correction of intensity compression, erasure of the imaging system, unwrapping of the spherical images, filling-in blank regions, and stitching images collected from different angles. We applied our method to Wytham Woods, an ancient semi-natural woodland near Oxford, UK. We acquired a total of 168 images in two sites with low and high abundance of ash, leading to differences in canopy, leading to a total 14 hyperspectral light probes. Our image-processing pipeline corrected small (<3 °) field-based misalignment adequately. Our novel hyperspectral imaging method is adapted for field conditions and opens up novel opportunities for capturing the complex and dynamic nature of the light environment.