Automatic multi projector camera calibration using fiducial marker detection

Abstract

This thesis introduces a system that enables projection mapping of multiple uncalibrated projectors and cameras onto a surface of any shape, without the need for specialized expertise or expensive equipment. Projection mapping is a technology with numerous applications, including virtual object augmentation and spatial experience extension. Achieving precise alignment of multiple uncalibrated projectors and cameras traditionally requires additional information, such as physical patterns, pre-calibrations, or 3D models. This process is both time-consuming and expensive. The proposed method utilizes fully automated geometric correction that does not require any additional information and can function under varying lighting conditions in the room. The process involves projecting a marker pattern onto a surface, capturing it with a camera, and using a marker estimation network to detect the markers and estimate their coordinates. This creates a 2D correspondence between the camera and projector, enabling the establishment of a pixel relationship between each camera-view image and projection-view image. This system can compensate for geometric distortions in camera views during projection mapping. This technology enables beginners to utilize projection mapping without requiring expert knowledge. It can be configured with multiple low-cost web cameras and general projectors. This system has a significant impact on the utilization of projection mapping in diverse applications, including entertainment, advertising, and education. This represents a significant advancement in performing projection mapping using multiple uncalibrated projectors and cameras, making it more efficient and accessible to a wider range of users. The proposed system is likely to drive new advances in projection mapping technology, and its potential applications are numerous.