Research Progress of Deep Learning-Based Wood Surface Defect Detection

Yuan Zhang; Xuewen Ding

doi:10.6911/WSRJ.202510_11(10).0008

Authors

Yuan Zhang
Xuewen Ding

DOI:

https://doi.org/10.6911/WSRJ.202510_11(10).0008

Keywords:

Wood surface defects; deep learning; object detection; attention mechanism; multi-scale feature fusion.

Abstract

The detection of surface defects in wood is crucial for enhancing product value and manufacturing efficiency. Traditional methods exhibit inherent limitations in both efficiency and robustness, whereas deep learning techniques offer innovative solutions to these challenges. This paper provides a comprehensive review of the research progress in this field, beginning with an overview of the evolution from convolutional neural networks (CNNs) to object detection and image segmentation models, as well as their adaptability in wood defect identification. It then focuses on the role of key optimization strategies, such as attention mechanisms and multi-scale feature fusion, in improving detection performance under complex textures. Finally, this paper highlights future directions, including the integration of multimodal information, the development of self-supervised learning, and the construction of end-to-end systems, all aimed at providing forward-looking technological references for the intelligent upgrading of the wood industry.

Downloads

Download data is not yet available.

References

[1] Future projections of global wood carbon flow and stock and climate change mitigation potential considering waste wood recycling[J]. Biomass Bioenergy, 2025.

[2] Song L, Liu Y T. Research on weighted HOG and feature fusion behavior recognition method[J]. Software Guide, 2021, 20(11): 53-56.

[3] Wang X F, Liu C, Zhao Y Q. Threshold automatic selection algorithm based on grey system theory[J]. Computer Applications, 2010, 46(19): 154-156+219.

[4] Xiao Y Q, Yang H M. Research on application of object detection algorithm in traffic scene[J]. Computer Engineering and Applications, 2021, 57(6): 30-41.

[5] Xiao J Y, Ma Y, Zhou Z, et al. Wood surface knot defect detection based on Faster R-CNN model[J]. Wood Industry, 2020, 34(02): 45-48.

[6] Zhu Y F. Research on wood surface defect detection based on Transformer[D]. Liaoning University of Science and Technology, 2024.

[7] Xia J J. Research on wood surface defect detection based on machine vision[D]. Qilu University of Technology, 2025.

[8] Ansari M M, Kumar S, Bin Heyat M D, et al. SVMVGGNet-16: A novel machine and deep learning-based approach for lung cancer detection using combined SVM and VGGNet-16[J]. Current Medical Imaging, 2025, 21: e15734056348824.

[9] Gao M Y. Research on wood knot defect detection based on ResNet convolutional neural network[D]. Northeast Forestry University, 2022.

[10] Dong H S, Xian X F, Sun X, et al. Tea disease recognition based on multi-stage feature fusion MobileNet v3-ULAM[J]. Jiangsu Agricultural Sciences, 2025, 53(15): 200-211.

[11] Zhao Q, Qu J W. An improved SAM-ShuffleNet Ming Qinghua dragon pattern chronology model[J]. Modern Electronic Technology, 2025, 48(09): 116-123.

[12] Rong Q, Tian Q C, Tan R. Real-time wood surface defect detection method based on improved Yolov5s[J]. Forest Products Industry, 2025, 62(01): 64-71.

[13] Ge Y, Ji H, Liu X. Wood surface defect detection based on improved YOLOv8[J]. Signal, Image and Video Processing, 2025, 19: 663.

[14] Zhang C, Zhang S. A lightweight vehicle information detection method for bridges based on semantic segmentation and two-branch neural networks of monocular vision[J]. Signal, Image and Video Processing, 2025, 19: 1032.

[15] Deng Z C, Yang Z C, Chen C, et al. Multimodal plane instance segmentation with the Segment Anything Model[J]. Automation in Construction, 2025, 180: 106541.