Geometric calibration of line-scan camera using a planar pattern

Abstract

We present an innovative calibration method for line-scan cameras to estimate the intrinsic parameters. The calibration involves using a stationary planar pattern that consists of repeated vertical and slanted lines, and constructing a two-dimensional (2-D) calibration framework with one-dimensional (1-D) data. A feature point reconstruction method is applied to transform the 1-D camera calibration problem into the 2-D scope. Camera parameters are then solved by using a 2-D camera model with constraints unique to 1-D geometry. In our tests over 12 calibrations with images of 2048×2048 pixels , the average of the reprojection errors is 0.46 pixels. As opposed to other line-scan camera calibration techniques, this method does not require the camera to progressively scan a pattern, thus eliminating the need for additional mechanical devices to assist the calibration. This method does not need a three-dimensional pattern as a calibration target, either. The stationary planar target makes the calibration more suitable for an application that has to be done in a nonlaboratory setting, such as highway pavement inspection.