Lecturer: Sota Nanjo (Tokyo) Date: 2023-03-16 10:00 Place: Aniara
Auroral observations using a wide-angle digital camera mounted on a gimbal
University of Electro-Communications, Tokyo
The seminar is also online. See emails for connection details.
We present the results of an auroral observation in Kiruna, Sweden, from November 2022 to March 2023 using a commercial wide-angle (horizontal field of view of ~73 degrees) digital camera. An observation with a wide-angle lens has the advantage of increasing spatial resolution but has the disadvantage of a narrow field of view (FOV). In this study, we have installed the camera on the gimbal to point the camera in the direction of the aurora to overcome this limitation. In our method, the direction of the appearance of the aurora is derived by referring to the real-time all-sky image by another digital camera, which is in operation at the same observatory in Kiruna. Specifically, the all-sky image is divided into 49 smaller sub-images, and each image is classified by a deep learning model ResNet-50 (He et al., 2016), to determine the regions of aurora. Based on the classification result, the camera is pointed in the direction of the aurora by operating the gimbal. When auroras are detected in multiple regions, the region with the largest elevation angle is selected. In order to capture fast auroral modulations, such as the internal modulation of pulsating auroras, video recording is carried out at 30 frames per second instead of still imaging. Each sequence of the recorded video has a length of 60 minutes maximum, but it would be shortened if the aurora disappeared within 60 minutes.
When analyzing the captured video data, it is necessary to remember that the transmission characteristics of the RGB channels are unknown and that the raw counts before software processing were not saved (data could not have linearity to the light intensity). Optical calibration experiments were conducted at the National Institute of Polar Research prior to the observation to evaluate these effects. We derived the wavelength characteristics of the RGB channels using a monochromator and confirmed that the linearity could be reconstructed from the recorded data by using the "S-log 3" picture profile, in which the function to compress raw counts to RGB values is publicly available.
Created 2023-01-30 14:48:12 by Mats HolmstrÃ¶m Last changed 2023-03-15 09:52:12 by Mats HolmstrÃ¶m