A user-friendly method to geometrically calibrate projector-camera systems

Abstract

Projector-camera systems drive applications in many fields such as measurement and spatial augmented reality. When needed, we can find their internal and external parameters via geometric calibration. For this process, we have to use both a printed pattern and a projector pattern, but they can easily interfere with each other. Current methods compensate by decoupling their calibrations or by leveraging structured light and color channels, but the required manipulations are not user-friendly. Therefore, we cannot expect normal users to execute the procedure, which can also become a burden for researchers. Although not always required, knowledge of the geometric parameters can often facilitate development of new systems. To make the calibration process easier, we propose a method that uses fiducial markers, from which we can easily derive a prewarp that, once applied to the projector calibration pattern, prevents its interference. Using our method, we confirmed that users can easily calibrate a projector-camera system in less than one minute, which we consider to be user-friendly, while still achieving typical subpixel accuracy.

inFORM

“…The projector's world coordinates were determined after calibration with the Kinect color camera, while distortion coefficients and intrinsic parameters were recovered using ProCamCalib [1]. The projector's 1400×1050 pixel output is then warped to correct for the parameters and projected over an 87×66 cm area.…”

Geometric calibration for a multi-camera-projector system

“…To decode each bit in the projected binary 1 An alternative to direct binary coding for the binary patterns is Gray codes. We have empirically found direct encoding of the column indices to be sufficient since we only keep the decoded positions for pixels that are confidently decoded.…”

“…This process is repeated for all cameras observing the binary projected patterns. 1 We opt to use a translucent, planar sheet as the projection target to calibrate all sensors. Translucency is needed so that even though the target is being projected from one side, the cameras viewing it from the opposite side can still view and decode the projected patterns.…”

“…Projector calibration has also been studied extensively. Since projectors cannot observe the scene, projector calibration has traditionally been preceded by camera calibration [20,12,1]. Alternatively, it is possible to calibrate a camera and projector simultaneously [3,18].…”

A Turntable-based Approach for Ground Truth Tracking Data Generation

“…There are algorithms where both projected and printed pattern are used (Audet, 2009;Martynov et al, 2011). The main idea here is that if the projected pattern is iteratively adjusted until it fully overlaps the printed pattern, then the projector parameters can be estimated.…”