Astrometric reduction of event camera observations

Abstract

Astrometric reduction turns an image of the sky into the angular position of an object, by recognising the field stars and solving the geometry of the exposure. Event cameras produce no images but streams of luminance changes, and on a fast moving object that is an advantage: the track arrives with very low latency and without saturating.

The thesis extends the BOPAS reduction pipeline to neuromorphic observations and characterises its accuracy against conventional sensors.

Objectives

1. Reconstruct usable positions starting from an event stream.
2. Integrate the reconstruction into the existing reduction pipeline.
3. Characterise the angular accuracy against observations with a conventional sensor.
4. Identify the motion regimes where the event sensor is genuinely worthwhile.

Methodology

1. Observations with both sensors on the same instrument, on the same passes.
2. Accumulation of events into time windows, with a study of the best window for star recognition.
3. Field recognition and astrometric solution with the existing pipeline.
4. Comparison of the derived positions with those from the conventional sensor, and with ephemerides for catalogued objects.
5. Error analysis as a function of the angular rate of the object.

Expected contributions

1. A working reduction pipeline for event based observations.
2. Accuracy characterisation against the conventional reference.
3. Identification of the angular rate threshold beyond which the event sensor is preferable.

Tools

Event camera and the lab optical instrumentation, the BOPAS pipeline, Python.