Wood Defect Detection Algorithm Integrating Multi-Scale Features
DOI:
https://doi.org/10.63313/CS.8024Keywords:
Wood Defect Detection, Multi-Scale, LightweightAbstract
This paper proposes a lightweight wood defect detection model, termed MADORE-Minify, which effectively addresses the limitations of existing methods in balancing multi-scale target representation and parameter redundancy in wood defect detection. The overall architecture of MADORE-Minify follows the YOLO11 framework and is further optimized and enhanced upon this baseline.
First, the backbone network integrates the Mixed Local Channel Attention (MLCA) module to strengthen feature extraction capability, enabling more effective detection of small-scale defects while al-leviating semantic information loss for large-scale targets. Second, the ADown downsampling module is employed to replace the upsampling layers, enhancing the model’s ability to capture features at different scales, particularly in preserving fine details of small targets. Meanwhile, this design reduces the spatial resolution of feature maps, thereby significantly decreasing computational cost and parameter complexity. Finally, an OREPA-GENet module is designed, which transforms nonlinear blocks into linear ones and com-presses complex structures into a single convolutional layer during the inference stage, achieving substan-tial model lightweighting and efficiency improvement.
Extensive experiments conducted on a public wood defect dataset demonstrate the superiority of the proposed MADORE-Minify algorithm. Compared with several competing methods, MADORE-Minify achieves a better balance between detection accuracy and model lightweighting. In addition, the proposed model attains high detection accuracy with only 6.6M parameters, validating its outstanding performance and lightweight characteristics in the field of wood defect detection.
References
[1] Hacefendiolu K , Ayas S , Baaa H B ,et al.Wood construction damage detection and localization using deep convolutional neural network with transfer learning[J].European Journal of Wood and Wood Products, 2022. DOI:10.1007/s00107-022-01815-5.
[2] Shabani A , Kioumarsi M , Plevris V ,et al.Structural Vulnerability Assessment of Heritage Timber Buildings: A Methodological Proposal[J].Forests, 2020, 11(8):881. DOI:10.3390/f11080881.
[3] Hashim, Ummi Raba’ah, Hashim S Z M , Muda A K .Performance evaluation of multivariate texture descriptor for classification of timber defect[J].Optik - International Journal for Light and Electron Optics, 2016, 127(15):6071-6080. DOI:10.1016/j.ijleo.2016.04.005.
[4] Zhang S, Huang H, Huang Y, et al. A GA and SVM classification model for pine wilt disease detection using UAV-based hyperspectral imagery[J]. Applied Sciences, 2022, 12(13): 6676.
[5] Zhang Y , Liu S , Cao J ,et al.Wood board image processing based on dual-tree complex wavelet feature selection and compressed sensing[J].Wood Science & Technology, 2016, 50(2):297-311. DOI:10.1007/s00226-015-0776-y.
[6] Lecun Y , Bottou L .Gradient-based learning applied to document recognition[J].Proceedings of the IEEE, 1998, 86(11):2278-2324. DOI:10.1109/5.726791.
[7] Simonyan K , Zisserman A .Very Deep Convolutional Networks for Large-Scale Image Recognition[J].Computer Science, 2014. DOI:10.48550/arXiv.1409.1556.
[8] Lu Y, Chen Y, Zhao D, et al. Graph-FCN for image semantic segmentation[C]//International symposium on neural networks. Cham: Springer International Publishing, 2019: 97-105.
[9] Tan M, Pang R, Le Q V. Efficientdet: Scalable and efficient object detection[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 10781-10790.
[10] Wang L, Qin H, Zhou X, et al. R-YOLO: A robust object detector in adverse weather[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 72: 1-11.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 by author(s) and Erytis Publishing Limited.
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
How to Cite
