| dc.description.abstract | In the past, measuring the luminance distribution of the sky and the sun was done with either a sky scanner or a luminance meter. The conventional measurement methods are time consuming at low measurement resolution (with a maximum of 145 data points on the entire upper hemisphere), thus, cannot capture real-time changes typically seen with natural daylight. To solve this problem, a camera-array-based measurement technology was recently introduced by the University of Kansas lighting research laboratory to capture the spatial and temporal luminance distributions of the celestial hemisphere. This technique uses high dynamic range (HDR) photogrammetry for luminance mapping of the sky and the sun simultaneously. With two cameras mounted next to one another on a Sky Measurement Tripod Head developed in the lighting research laboratory, the sky and the sun are measured, respectively, by each camera. However, one issue that still remains with this type of data collection is the storage and treatment of big data embedded in the HDR images generated in the field. Each HDR image has a file size of approximately 40-50 MB, while the retrieved 18 million luminance data in text file format could have a file size of approximately 400-500 MB. Given at least hourly measurements for real-time sky conditions from sunrise to sunset, it is very tedious to deal with such large amounts of data that challenge the speed and storage capacity of current computation facilities. To solve this problem, the present research study was aimed to explore the feasibility of reducing pixel resolutions in the laboratory of raw HDR images taken in the field, in the hopes of speeding up the data treatment process of the sky and the sun luminance measurement while still maintaining an adequate degree of accuracy. An experiment was carried out at the Clinton State Park in Lawrence, KS at 1:30 pm on October 4th, 2015 to evaluate the null hypothesis that reducing the pixel resolution of the HDR images in the laboratory would not compromise the overall value of the obtained data. Two Canon digital cameras EOS Rebel T2i fitted with Sigma 4.5mm F2.8 EX DC HSM Circular Fisheye lenses were mounted side by side on a custom designed Sky Measurement Tripod Head to take measurements of the celestial sky using the HDR photography. One camera was mounted without a neutral density filter and was used to capture the luminance distribution of the sky while the other camera was equipped with a neutral density filter of 1/1000 and used to capture the luminance of the sun and its corona. The luminance data embedded in each of the two HDR images were later extracted in Radiance and outputted to Microsoft Access and Excel for the follow-up data treatment. It was discovered that the amount of data obtained from the cameras was very large and nearly impossible to handle in Microsoft Access or Excel due to their limited computation capacity of 18 million rows of data. This study then reduced such big data during the data extraction process in the laboratory by lowering the pixel resolutions of the raw HDR images obtained in the field. The size of the HDR images was reduced from 18 million data points to merely 270,500 data points. The reduced datasets were then treated using Excel spreadsheets containing pre-developed equations. Calibration Factor (CF) values were calculated by comparing the actual horizontal illuminance measured using an illuminance meter to the calculated illuminance from the sky and sun luminance data embedded in the synthesized HDR image. In theory, the CF ratio should be close to 1.0 indicating the robust data collection and treatment process was carried out with minimal error. In the present study, the CF value obtained during the laboratory data treatment was close to 0.05, indicating the dataset was improperly manipulated during the reduction process of pixel resolutions. Photometric calibrations using such a CF value (0.05) would lead to extraction of only 5% of the true luminance distributions of the sky and the sun. As a result, it is deemed inappropriate to reduce the pixel resolution of raw HDR images in the laboratory after the field measurement, since such a reduction found in this study is associated with a loss of useful data for luminance mapping of the sky and the sun. Further research to be conducted in the Lighting Research Lab will evaluate two possible ways to solve this problem. The first solution is to capture the HDR images with lower pixel resolutions by directly adjusting the camera settings in the field, which is not the optimal solution but recommended given the otherwise resulting big data and the limitations of current computing facilities. The second method is to conduct the data treatment in a more powerful computing software such as Matlab without reduction of the original 18 million pixels embedded in the HDR images. | |
