Bias Detection and Prediction of Mapping Errors in Camera Calibration

Abstract

Camera calibration is a prerequisite for many computer vision applications. While a good calibration can turn a camera into a measurement device, it can also deteriorate a system’s performance if not done correctly. In the recent past, there have been great efforts to simplify the calibration process. Yet, inspection and evaluation of calibration results typically still requires expert knowledge. In this work, we introduce two novel methods to capture the fundamental error sources in camera calibration: systematic errors (biases) and remaining uncertainty (variance). Importantly, the proposed methods do not require capturing additional images and are independent of the camera model. We evaluate the methods on simulated and real data and demonstrate how a state-of-the-art system for guided calibration can be improved. In combination, the methods allow novice users to perform camera calibration and verify both the accuracy and precision. Electronic supplementary material The online version of this chapter (10.1007/978-3-030-71278-5_3) contains supplementary material, which is available to authorized users.

Inferring bias and uncertainty in camera calibration

“…As a generic choice, we have used the average error across the image, but one could also weight image regions differently, or define an application-specific error, such as a triangulation error of a stereo system. Furthermore, the EME can be applied for calibration guidance, in order to guide users to collect calibration datasets that explicitly minimize the uncertainty [12].…”

“…In general, the error sources of camera calibration can be divided into systematic errors (bias) and uncertainty (variance) [10, p.116] [12]. As in any data modeling problem, systematic errors occur if a chosen model is not sufficiently flexible or imposes false assumptions on the data (Fig.…”

“…Our goal is to reliably detect and quantify both types of errors and to condense the result to easily interpretable measures. Extending our recent work on evaluation metrics [12], we provide four main contributions (Fig. 1):…”

Modeling dynamic target deformation in camera calibration

“…To compare the different sets of intrinsics, we used mapping error in image space [3,5,6]. This metric propagates the differences in the individual parameters to a difference in image space.…”

“…To express the deviation in the individual parameters as a single number, we computed the mapping error w.r.t. the reference calibration [5,3,6]. Fig.…”