Multispectral image correction for geometric measurements

Multispectral- and hyperspectral imaging technologies enable new possibilities in industrial measurement applications. Based on the knowledge of remote sensing a lot of investigations were made in the last decades of years. Nevertheless the demands on remote sensing versus technical multi spectral image processing are quite different. In the field of precise geometric measurement technics it is necessary to correct the image data between different spectral channels with a high accuracy, normally in the micron range. Otherwise the geometric absolute value of fail detection on edges can be become very large. State of the art in industrial imaging and detection of geometric features is the calibration of only one imaging channel. In this paper, the studies on a twelve channel multi spectral imager were presented. For the applied filter wheel system, investigations on the improvement of lens aberration as well as for the defocus problem were made. Therefore a calibrated high precision geometric test chart was used to calibrate the system geometrically. To correct the geometric errors on the image plane a special moving filter approach, based on linear convolution, was developed. For every channel a calibration matrix were calculated and applied on the image system output.