Computer vision model for recognizing bark in a digital image of a tree cross-cut

The aim of the study is refining the algorithm for binarization of digital images of cross-cut round timber for automated assessment of the bark ratio, as well as testing methodology for adjusting its parameters. The program implementation of the refined algorithm and calculations are performed in Python, the main functions used are implemented in the OpenCV library. The image binarization algorithm used to assess the bark ratio bases on the single threshold method. The method parameter settings are made for the author's images of cross-cut round timber. To obtain experimental data, cross-cut round timber was photographed from different angles. The sample intended for testing the proposed algorithm and software solution included 130 images of crosscut timber with a diameter of 20–30 cm; wood species – birch, spruce, alder. The photographing was performed in winter conditions, on a snow-covered background under natural light in the absence of direct sunlight. The resolution of the original images was 3024x4032 pixels (72 dpi). The algorithm proposed includes adjustment of the parameters of the single threshold binarization method basing on information about the photosensitivity, the brightness level of the image, and the exposure value; the method is called twice during the image processing. Heuristic dependencies are given for determining the threshold values of the single threshold method parameters for its two iterations performed sequentially when separating an object from the background and further segmenting the bark in the image. Examples of the results of determining the ratio of bark in images with an uneven contour and a dominant background, with a relatively smooth contour of a cross-cut in the image, as well as with a partially icy surface of a cut with an uneven contour, as well as a link to a repository with experimental images and the program solution implementing the proposed algorithm are given. Testing the method for adjusting the parameters and the algorithm taking into account various characteristics of the image and the object in it showed that the solution allows obtaining satisfactory results.

1. Beliaev N.L. Povyshenie tochnosti tsifrovogo ucheta kruglykh lesomaterialov: Diss. … Candidate of Technical Sciences. Petrozavodsk: PetrSU, 2023. 163 p. (In Russ.)

2. Bogomolov A.A., Sukhov A.S., Trubnikov A.V., Khitrov E.G. Instrumenty biblioteki OpenCV dlia vyiavleniia kory na tortsevom snimke poperechnogo raspila dereva. Sovremennoe programmirovanie: materialy mezhdunarodnoi nauchnoprakticheskoi konferentsii Nizhnevartovsk, 2024, pp. 412–416. (In Russ.)

3. Bogomolov A.A. Repozitorii programmy dlia avtomaticheskogo opredeleniia doli kory na tsifrovom snimke tortsa poperechnogo raspila stvola dereva, 2025. URL: https://clck.ru/3LreKK (In Russ.)

4. Han S., Jiang X., Wu Z. An Improved YOLOv5 Algorithm for Wood Defect Detection Based on Attention. IEEE Access, 2023, vol. 11, pp. 71800–71810. DOI: 10.1109/ACCESS.2023.3293864.

5. Kim T.K., Hong J., Ryu D., Kim S., Byeon S.Y., Huh W., Kim K., Baek G.H., Kim H.S. Identifying and extracting bark key features of 42 tree species using convolutional neural networks and class activation mapping. Sci Rep, 2022, vol. 12, art. no. 4772. DOI: 10.1038/s41598-022-08571-9.

6. Kunickaya O., Pomiguev A., Kruchinin I., Storodubtseva T., Voronova A., Levushkin D., Borisov V., Ivanov V. Analysis of modern wood processing techniques in timber terminals. Central European Forestry Journal, 2022, vol. 68(1), pp. 51–59.

7. Robert M., Dallaire P., Giguère Ph. Tree bark re-identification using a deeplearning feature descriptor. 17th Conference on Computer and Robot Vision (CRV). 2020, pp. 25–32. DOI: 10.1109/CRV50864.2020.00012.

8. Wang R., Chen Y., Liang F., Wang B., Mou X., Zhang G. BPN-YOLO: A Novel Method for Wood Defect Detection Based on YOLOv7. Forests, 2024, vol. 15, iss. 7, art. no. 1096. DOI: 10.3390/f15071096.