CALIBRATING BINOCULAR VISION SENSOR WITH ONE-DIMENSIONAL TARGET OF UNKNOWN MOTION

Abstract

The mathematical model of the binocular vision sensor is analyzed. A flexible new approach to easily determine the geometric parameters of binocular vision sensor with one dimensional target of unknown motion is presented. The technique only requires the sensor to observe a one-dimensional target shown at a few (at least four) different orientations after the sensor has been fixed. All the accurate geometrical parameters of the binocular vision sensor are estimated by nonlinear minimization of the cost function relative to the distance error affected by the measurement errors of the three dimensional coordinates of the spatial point and the distance errors between the projective point and the observed points of the same spatial point. The proposed method is simple and need not to provide initial parameters. Moreover, the calibration status of the binocular vision sensor is identical to that of the work status. The experimental results prove that this method is practical and is accurate to within 0.03 nun when the measuring range is about 200 mm × 200 mm.