Discrimination of bark from wood chips through texture analysis by image processing

Utilization of wood chips for bioenergy requires classification and segregation of the constituents of the chipped mass to help optimize energy conversion. Wood chips obtained from processes such as forest thinning can contain a considerable amount of material other than wood chips, such as bark. An image processing algorithm was developed to discriminate bark from wood chips. The algorithm involved object identification, image capture, single value decomposition to describe wood texture evident in grayscale image with a single numerical value, and application of logistic models involving the single values representative of wood texture to predict whether a chip is bark. The percentage of correct predictions using this system was about 98%.

Graphical AbstractFigure optionsDownload as PowerPoint slideHighlights
► An image processing system was developed to discriminate bark from wood chips.
► Wood texture represented by image parameters keyed bark–wood discrimination.
► Multiple image parameters were condensed through single value decomposition.
► A developed singular value, ratio of angles, produced the best discrimination.
► Logistic models with single values representing wood texture had 98% successful discrimination.