Abstract:

In this paper, we present a new camera calibration approach by taking three images at least of the proposed planar pattern which includes an arbitrary circle with its inscribed regular triangle under different orientations. First, the imaged circular points are derived from the images of the proposed planar pattern. And then, the camera intrinsic matrix and extrinsic matrix are determined via the obtained imaged circular points. In contrast to the conventional methods, our method avoids ellipse and line fittings. It reduces the computing complexity and it is more robust to noise. Another advantage of our method is that neither any metric measurement on the model plane, nor the point correspondences is necessary; hence the calibration process becomes extremely simple. Experiments on both synthetic data and real image data demonstrate the robustness and effectiveness of our method.

Key words: circular point;imaged absolute conic;camera calibration