Automatic grading of softwood sawlogs for pricing using external geometry

Abstract

External geometric factors obtained from three-dimensional (3D) log scanners can be used to sort logs according to different quality standards. This study evaluated the possibilities of automatic grading of logs according to current Swedish grading standard for pricing. The accuracy of the automatic method was compared with the accuracy of conventional manual grading. Models were developed from a training set using variables enhanced from a 3D log scanner and linear discriminate analysis. The models were tested on a large test set for which the grades set from both conventional manual grading and a grading inspector were known before the study; the latter was considered the true grade. The test set consisted of pine and spruce logs from five sawmills in different parts of Sweden. The accuracy of the grading was evaluated with Cohen's kappa. The evaluation compared the accuracy of manual versus automatic grading: this comparison was conducted in various ways. Automatic grading was determined as less accurate than manual grading for all logs. Although the accuracy of automatic grading was lower, advantages such as reduced costs could make automatic grading a preferable method of grading logs.