Calibration of Multiple Non-Gimbal Mounted Camera-Mirror Systems

Abstract

The primary objective of this Master’s thesis is to investigate methods for calibrating multiple sensor rigs designed to track the position of an object in 3D space. Each rig consists of a statically mounted sensor directed downward at a controllable pantilt mirror, allowing for adjustment of the sensor’s field of view. The calibration method compensates for misalignments both within individual rigs and between different rigs, and enables all rigs to observe the same 3D point. This thesis project focuses on calibrating a setup consisting of two rigs.

Based on a mathematical model of the system and measurement data collected with the sensor rigs, the positions and orientations of the rig components, which define the system parameters, are estimated. Once these parameters are known, the orientation required for one of the mirrors can be predicted so that its corresponding sensor observes the same 3D point as the other sensor. The parameter estimation is performed by minimizing a cost function, and two optimization methods have been evaluated for this purpose. The proposed calibration method has been tested on both simulated and real-world systems.

The performance of the parameter estimation was assessed in several aspects, including accuracy dependence on the number of calibration points, approximate angular errors in various calibration and test scenarios for both simulated and real-world systems, and some investigation of the sensitivity of the parameter estimation to variations in initial parameter estimates. Results show promising accuracy in some cases, with real-world tests indicating angular errors within ±0.3◦. However, a key limitation of the current method is its sensitivity to the initial parameter guess, which must be close to the true values for accurate parameter estimation. Addressing this limitation could improve the method’s applicability.