Normal Line Invariance-Based Lens Distortion Calibration Applied in a Precision-Sensitive Self-Calibrated Planar Metrology System

Although the decoupled straight line-based method is more flexible than full metric calibration, precision is the main burden for its practical use. Therefore, this paper introduces a novel method to estimate an optimal lens distortion model using local features of straight lines from the image, which are easily found or established patterns. The proposed method exploits the geometric invariance of the normal line passing the center of distortion (COD) for accurate normal vector estimation. The grid search technique within a narrowed search range determines an optimal COD position at a reasonable calculation cost. The metric of line straightness after undistortion is used for performance evaluation. This quantitively proves that the proposed method has superior undistortion performance than the state-of-the-art straight line-based lens and metric camera calibration methods. Combined with a simple homography estimation, the method is verified in a precision-sensitive self-calibrated planar metrology system. The system is calibrated with a few target objects, one of which has known geometry. Experiments show that the calibration process for the system achieves competitive measurement accuracy compared with state-of-the-art full metric calibration while being more flexible.