Abstract

We are dealing with a problem of identifying streaks detected by a telescope of an earth orbiting object. The problem is reformulated into a clustering problem. A theoretical study is performed to show that the hierarchical algorithm fits the problem better then the k-means algorithm. The theory is tested through a series of experiments using Matlab routines for hierarchical clustering. The experiments result in conclusions that there needs to be theory created for choosing the cut-off parameter for the algorithm. Finally a section method is introduced for a future development of computationally efficient algorithms for large cardinality of the problem. The work completed gives a direction into what part s of the hierarchical algorithm need to be

Citations