Automated Interpretability Scoring of Ground-Based Observations of LEO Objects with Deep Learning

The Space-object National Imagery Interpretability Rating Scale (SNIIRS) allows human analysts to provide a quantitative score of image quality based on identification of target features. It is naturally difficult to automate this scoring process, not only because the scale is based on identifiable features but also because the images may be in an almost-resolved image quality regime that is difficult to handle for traditional machine vision techniques. In this paper we explore using a convolutional neural network to automatically produce SNIIRS scores. We use wave-optics simulation with varied turbulence strength to generate a dataset of images of Low-Earth Orbit (LEO) satellites observed from a ground-based optical observatory. SNIIRS scores are automatically generated for these images based on a combination of a priori knowledge of each object's simulated features and the simulated turbulence strength. A neural network is then trained to provide accurate SNIIRS scores from single images without being provided knowledge of the object model.