Accession Number : AD1001967


Title :   RSO Characterization with Photometric Data Using Machine Learning


Descriptive Note : Conference Paper


Corporate Author : Charles River Analytics, Inc Cambridge United States


Personal Author(s) : Howard,Michael ; Klem,Bernie ; Gorman,Joe


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/1001967.pdf


Report Date : 18 Oct 2015


Pagination or Media Count : 10


Abstract : Object characterization is the description of a resident space object (RSO), its capabilities, and its behavior. This paper explores object characterization methods using photometric data. An important property of RSO photometric signatures is the changes in intensity that they exhibit with respect to changes in viewing angle or orientation. Properties that influence the brightness of the photometric signature include geometry, orientation, material characteristics and stability. For this reason, it should be possible to recover these characteristics by analyzing photometric signatures. In this paper, we discuss the application of machine learning techniques to RSO characterization. We develop simulated signatures in the visible band of three basic RSO types, with variations in object orientation, material characteristics, size and attitude. We generate observations by sampling measurements from the simulated signature. Next, we apply a feature extraction technique to the simulated signatures and train a variety of machine learning algorithms to classify the signatures. The classifications are made on sequences of one or more measurements. We consider the effectiveness of a set of binary classifiers trained to recognize one of each case. The results of each classifier are combined together to produce a final output characterization of an input observation. Experiments with varying levels of noise are presented, and we evaluate models with respect to classification accuracy and other criteria.


Descriptors :   machine learning , space objects , photometry , computer vision , communication satellites


Distribution Statement : APPROVED FOR PUBLIC RELEASE