Intrinsic Parameter Calibration of Line-Scan Cameras Using RANSAC Algorithm

In the calibration process of line-scan cameras, the feature points are usually scant and apt to be interfered by noise. Therefore, an innovated method with high precision for calibrating the line-scan camera parameters is proposed in this paper. Firstly, a 2-D calibration pattern is placed in different positions of the measured field, and a large number of feature points are collected. And then, in order to reduce the effect of the errors, an iteration solution method based on the RANSAC (Random Sample Consensus) algorithm is put forward. The parameter model with minimum reprojection error is chosen to obtain intrinsic parameters with high precision. The experiments show the average of the reprojection error using this method is about 0.3594 pixels. Compared with the current methods, the calibration method proposed in this paper does not require the use of the specially designed 3-D calibration pattern and the assistance of additional mechanical devices. This method is proved to be high precision and suitable for practical applications.