An automated image-quality characterization pipeline for industrial camera performance evaluation

Industrial barcode scanners are usually judged by how many codes they read, which says little about the quality of the images behind those reads. This thesis presents an automated pipeline that measures image quality directly, using pictures captured by a Matrix 930 scanner mounted over a conveyor at a 45° viewing angle. A custom target carrying ArUco markers was designed and printed in three sizes, fixed to boxes, and photographed as they travelled under the camera. For each image the software detects the markers, identifies which target is in view, computes a homography that maps the regions of interest from a JSON layout definition onto the real image, and then measures SNR, CNR, contrast and MTF50. Results are averaged over the whole dataset and grouped by grid region so that performance can be compared across the field of view, then exported as an interactive HTML report. The outcome is a repeatable procedure that replaces slow manual inspection and makes camera comparisons consistent.