Bark Abrasion Evaluation on Birch (betula Spp.) Round Timber by Using Machine Learning Algorithm
Year: 2025
Type: conference-paper
Abstract: Birch has an exceedingly high potential from the point of view of cultivation, respectively, now in Latvia more and more new factories are being opened that produce birch plywood and the demand for birch round timber is increasing every year. Latvian legislation, as well as business relations between different trading parties, require accurate determination of the volume of round timber without bark. Automatic measuring devices mainly determine the diameter with bark and then reduce it using bark algorithms. Machine learning has great potential in labour shortage. During machine learning, a computerized system could analyse large amounts of data and diverse properties [1]. Convolutional neural network can ensure the system's resistance to image defects due to various types of lighting conditions, image shifts and changes in their shapes, which can be caused by the characteristics of the camera lens [2]. One of the tasks of timber measurement process is to assess the amount of bark abrasion, as well as the thickness of the bark. Within the framework of the project, we have investigated the possibilities of determining the area of bark abrasion of round timber using machine learning algorithms. To assess the accuracy of the model, we have randomly selected 90 round timber samples from the system, which have been marked by a timber scaler in a computer program, as well as the computer program itself. On software with image processing and analysis capabilities, visual data of the side surface of round timber were evaluated, manually assessing the areas of bark abrasion. The obtained results were compared with the bark mass obtained using the prepared computer vision algorithm. Machine learning algorithm, on average, calculates smaller bark mass – 76 %, compared to the manually obtained results – 84 %. Bark abrasions and most of the remaining bark can be assessed without difficulty for round timber with a darker and more crusty bark, for example, for cuts prepared from the thick end of the trunk. For birch round timber with a white colour, in some cases, part of the white colour is perceived as bark abrasion and, therefore, a smaller majority of remaining bark is assessed. To assess the volume of round timber without bark, the proportion of bark must be assessed, which must be calculated from the total volume of the measured diameter with bark. When assessing the coincidence of the determined bark types, it can be observed that the bark type according to the computer vision algorithm coincides in 79 % of cases with the bark type obtained after manual processing of visual data.
Authors: Janis Magaznieks, Mareks Millers
Language: English
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4411251599