Multiple Projector Camera Calibration by Fiducial Marker Detection

Abstract

Projection mapping has been used for various purposes in everyday situations. A key step in projection mapping is to project images and videos without distortion through calibration, which is typically performed manually. Calibration becomes more challenging when multiple projectors are involved. To address this issue, a fully automated calibration method for a multi-projector-camera system is proposed in this paper. The projectors and cameras are assumed to be un-calibrated, and an arbitrary geometric shape of the projection surface is considered. Without using checkboards or user-provided parameters, the proposed method can automatically estimate calibration parameters for the cameras and projectors and generate compensated content for projection without distortion within a reasonable amount of time. The proposed method utilizes AprilTag markers and modified YOLOv8 with deformable convolution for robust marker detection and correspondence estimation between the projectors and cameras, providing an automatic process for completing calibration and distortion correction. Various experiments have demonstrated that the proposed method outperforms existing methods using checkerboards in terms of calibration accuracy and processing time across various camera-projector configurations. The proposed method can minimize the difficulty of projection mapping, allowing it to be used in everyday situations without requiring a certain level of knowledge about projection mapping theory and related hardware.