多通道投影图像的几何自动拼接技术

Automatic geometry alignment for muti-projector image display wall

以低端摄像头产品为测量元件构造了实用的可扩展的多通道投影图像自动拼接系统;利用二次校准方法细化求解一致性矩阵,在投影屏幕上实现了多通道投影图像间的几何无缝拼接;特征图案的正确选择和二次校准方法最大限度地降低了摄像头性能对校准精度的影响;在满足相邻投影图像有重叠区域和摄像头能拍摄到重叠区域两个基本条件时,系统可实现自动拼接校准.实验表明,二次校准方法使校准时间明显缩短,使校准精度达到了亚像素级;低端摄像头、高效率的特征提取方法和一致性矩阵求解算法使系统扩展成本大大降低.

A practical and scalable alignment automation system for muti-projector display wall was constructed based on the measurement of low-performance cameras. Taking the advantage of re-alignment theory the homography matrix was refined and used to achieve seamless alignment across different areas on the projection screen. The special pattern for feature extraction and the re-alignment idea furthest decreased the effect of the cameras＇ performances on the alignment precision. The system can automatically align the projectors under the conditions that there are overlapped view areas between adjacent projectors and cameras can observe them. It is shown in the experiments that the re-alignment method obviously shortens the calibration period and brings us sub-pixel alignment precision. The low-performance cameras, high efficient feature extraction method and sophisticated solution for homography make the cost of expanding the system low.