Master's Thesis Defense

Kevin Zhang

(Faculty Advisor: John Christian)

"Optimized Star Field Camera Calibration for Precision Optical Navigation"

Friday, April 10

3:00 - 5:00 p.m.

Coda C1215

Abstract: 

Calibrated cameras are essential for the use of vision navigation algorithms. Calibration of spacecraft optical instruments typically occur preflight, however mission experience has shown that calibration parameters change on orbit. Consequently, in-flight geometric camera calibration is a fundamental capability where precision navigation is demanded. Practically every mission relying on optical navigation calibrates cameras with star field images, opportunistically captured as mission operations allow. This practice neglects the quality of images collected, which is experimentally confirmed to yield less precise and stable calibrations. The non-uniform distribution of known stars can be exploited to maximize information provided in calibration images. An optimal design of experiments for actively sequencing star field images is formulated while considering operational pointing constraints. This principled approach minimizes time and data volume resources associated with capturing larger image sets.

Calibrating cameras becomes cumbersome at wider field-of-views (FOV). Wider FOV cameras enable powerful methods such as terrain relative navigation onboard landers, but they suffer from severe, complex distortions. A pinhole camera model can be simultaneously estimated with attitude in the well-studied lost-in-space problem. Unfortunately, the pinhole model is a linear model only sufficient in characterizing narrow FOV cameras (typically less than 30 degrees) and must be complemented with a nonlinear lens distortion model. Current methods of characterizing and calibrating lens distortion tend to be manual and error prone. An automated framework is proposed to perform robust global star identification using coherent point drift registration. Then, a data-driven distortion model parameterization is selected using cross validation. These contributions reduce mission operator burden and ultimately minimize biases performing spacecraft optical navigation.

Committee:

Dr. John Christian (advisor), School of Aerospace Engineering
Dr. E. Glenn Lightsey, School of Aerospace Engineering
Dr. Koki Ho, School of Aerospace Engineering
Courtney Mario, Draper