Calibration of multimodal 3D structured-light systems using digital features

Abstract

Calibration of multimodal 3D imaging systems that combine structured light with an additional modality typically relies on targets constructed with physical features that must be detectable by all imaging modalities. Such targets can be costly to produce and are prone to fabrication defects that degrade accuracy. Furthermore, reflections, light saturation, and the limited resolution of non-visible-range cameras complicate reliable feature detection. We present a calibration approach that uses digital features generated by a screen, a mirror, and an auxiliary camera—removing the need for specialized targets with physical features. This setup recovers the intrinsic parameters of the visible camera as well as the intrinsic and extrinsic parameters of both the projector and the additional-modality camera. To illustrate our method, we employ a thermal camera, though the procedure extends readily to other imaging modalities. Experimental results show that the proposed solution achieves a 0.07 mm root-mean-square error in 3D reconstructions, matching conventional techniques. By eliminating the requirement for physical features for targets, this approach reduces costs, avoids fabrication flaws, and simplifies multimodal feature detection.