Pattern recognition as tools for understanding night sky brightness variabilities

Night sky brightness (NSB) research related to the artificial light pollution issues has been increasing all over the world, using various measurement techniques and tools. The research produces tens of thousands of data for each month such that proper handling, processing and analysing the data become challenging. In this article, we demonstrate an alternative method for processing the NSB data by utilising pattern recognition techniques: Canny edge detection and Hough transform. These techniques were applied to identify data and extract important parameter from the NSB density plot semi-automatically. Datasets collected from Bandung, Garut, Subang and Sumedang used as the test cases. Three time-segments (dusk, night and dawn) became the main focus of the analyses and our method successfully extracted following parameters: the rate of sky brightness change at dusk and dawn, the average NSB at night and the intersections which indicate transition time. This method, along with its many possible improvements, enables us to process data more effectively and encourage more observation campaigns to be conducted in the future.