Accurate Sensor Calibration for Geometric Computer Vision

Abstract

Calibration of devices is the process of determining some internal or external parameters of a selected mathematical model for a specific device. This dissertation focuses on the calibration of widely used modalities in computer vision, for example, structured-light scanners, cameras, and LiDAR (Light Detection and Ranging) sensors.Calibration is a crucial step for many applications in computer vision. A priori knowledge of parameters (e.g. the focal length or camera poses) makes the processing of other algorithms faster and more accurate. Some of these parameters can be estimated online, during the usage of the device. However, online algorithms achieve lower accuracy in general and only used when the device is unreachable by human and no manual calibration is available. The proposed methods in this dissertation focus on accuracy rather than automatic or object-free calibration.