Analysis of Block Matching Algorithms for the application of image mosaicing to online surface inspection of steel products

摘要

Steel industry requires high quality controls to ensure certain requirements of the final products. In on-line inspection applications for long steel products, a camera is usually placed in the top, in order to view a portion of the product while it passes under the sensor in a roll table. The high relation between the length of the steel product and the necessary resolution to carry out these quality controls generates a set of partially overlapped images, that in many cases must be superimposed in order to obtain a full large image of the target. The characteristics of the algorithms for this image stitching process are essential to incorporate these techniques into the production line, with specific requirements about accuracy, processing speed and robustness. This paper presents a comparison between several Block Matching Algorithms (BMA) applied to two radically different steel industry cases, in which the surface reconstruction is necessary for defects detection and measurements. A general method is outlined for both cases, and it's effectiveness is shown by means of experimental results.