Enhanced vision algorithm for concurrent assessment of Identification Marks for glass manufacturing

The aim of this paper is to present a robust and a reliable methodology based on concurrent evaluation of Identification Marks for Automotive Glass. The proposed methodology is modeled using Petri Nets and a test implementation of the algorithm was successfully achieved to evaluate a set of samples. The main contribution is an algorithm based on Mean Centroid Pose Correction (MCPC) of the segmented elements. The method heuristically evaluates combined outcomes of Binary Large Objects (BLOBs) segmentation, Edge Detection segmentation and Image Subtraction, which are then contrasted to get a reliable unique result. The result of the corrected MCPC segmented elements is evaluated to determine the presence and position deviation of identification marks on shaped glass. An important benefit of this methodology is that a cost effective implementation can be easily programed, a JAVA based application is presented to demonstrate that the proposed solution is more affordable than the use of Industrial Cameras with integrated processors and proprietary software, fast enough to get real time response and as reliable as commercial solutions.