Meteor cluster event indication in variable-length astronomical video sequences

Abstract

ABSTRACT In recent years, the study of parallel or cluster meteor events has become increasingly popular. Many imaging systems currently focus on meteor detection, but the algorithms exploiting the data from such systems do not investigate the probability of cluster or parallel meteor events. This paper presents a novel approach to indicate a potential meteor cluster or parallel meteor event based on variable-length astronomical video sequences. The presented algorithm consists of two main parts: meteor event pre-detection and meteor cluster event probability evaluation. The first part of the algorithm involves a meteor pre-detection method based on the Hough transform and the exact event location within the time domain. In addition to pre-detecting meteor events, the method outputs event trajectory parameters that are further exploited in a second part of the algorithm. This subsequent part of the algorithm then operates over these meteor trajectory parameters and indicates the probability of cluster occurrence. The algorithm is experimentally evaluated on video sequences generated by the Meteor Automatic Imager and Analyzer (MAIA) astronomical imaging system, covering the Draconid and September ϵ Perseid meteor showers. Compared to the current MAIA meteor detection software, the proposed part of the pre-detection algorithm shows promising results, especially the increased rate of correct meteor detection. The meteor cluster evaluation part of the algorithm then demonstrates its ability to successfully select related meteor event candidates (disintegrated from the same parental object) and reject unrelated ones.