Comparing Grade Classification Criteria for Automatic Sorting of Norway Spruce Saw Logs

Abstract

Grades of centre and side boards from 277 Norway spruce logs were combined to form binary response variables, here denoted as sorting criteria. Four different grading systems were tested. The log geometry variables unevenness, butt taper and top taper were used in logistic regression models. The classification accuracy ranged from 58 to 83%. The accuracy was higher for visual stress grade criteria than for more complex criteria such as the Nordic timber grading rules. The number of tested criteria and thus possible comparisons limited the ability to establish significant differences. The low associations between board grades within logs and between graders, highlight key issues when developing and improving automatic log sorting systems.