Neural Network Classifiers to Grade Parts Based on Surface Defects with Spatial Dependencies

Abstract

In many manufacturing operations, unfinished or unassembled parts are not necessarily accepted or rejected in their entirety. Oftentimes, they are given some qualifying grade that indicates their worth to potential buyers or their suitability for particular processing operations. The manufacture of hardwood lumber in sawmills follows this general procedure. Lumber is graded based on appearance because its end use is primarily furniture and other goods that have a large aesthetic aspect to their value. Lumber grades are used by sawmill operators (sellers) and furniture plants (buyers) as an estimate of how much clear wood material is in the lumber being traded, hence the lumber’s value. Also, certain sawmill operations treat a board differently based on actual or potential lumber grade. Therefore, accurate lumber grading is important to mill operators.