Classification of wool and cashmere fiber based on LBP and DWT features: performance comparison of random forest, AdaBoost, and KNN classifiers
Source abstract
Abstract
The automatic identification of cashmere and wool fibers presents a significant challenge due to their substantial structural and textural similarities. An image-based method is proposed in this article, which combines local binary pattern (LBP), discrete wavelet transform (DWT), and several classification algorithms – including random forest (RF), adaptive boosting (AdaBoost), and K-nearest neighbors (KNN) – to achieve rapid and accurate differentiation between wool and cashmere fibers. Primarily, 300 images each of cashmere and wool fibers were captured using optical microscopy, and the original images were pre-processed to prepare input data for feature extraction. Texture features were then extracted using LBP, and energy entropy was extracted using DWT. Fiber diameters were measured through geometric morphological analysis, using the maximum inscribed circle method to determine fiber diameter from segmented images. In conclusion, these features were input into RF, AdaBoost, and KNN classifiers for model training and classification. Experimental results indicate that the RF classifier achieved superior performance, with a classification accuracy of 94.88%, outperforming both AdaBoost and KNN, and demonstrating high recall, and F1 scores. Overall, the proposed method exhibits exceptional accuracy, robustness, and generalization capability in fiber classification tasks, with significant potential for application to the classification of other natural fibers.
Source: openalex_inverted_index · Quality: baseline
Research context
OpenAlex topics: Textile materials and evaluations · Industrial Vision Systems and Defect Detection · Spectroscopy and Chemometric Analyses
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Textile and Fashion, Shanghai University of Engineering Science
Funding
National Natural Science Foundation of China
References (33)
Show source-provided references
1. 10.1016/j.compeleceng.2021.10719910.1016/j.compeleceng.2021.107199 ↗
2. 10.4028/www.scientific.net/AMR.821-822.24310.4028/www.scientific.net/AMR.821-822.243 ↗
3. Jiao, M. Y. (2017). A recognition method of cashmere and wool based on gray level co-occurrence matrix. Journal of Chengdu Textile College, 34(3), 126–129.(in Chinese with English abstract).
4. 10.1016/j.mehy.2020.10969610.1016/j.mehy.2020.109696 ↗
5. 10.1016/j.asoc.2015.06.00910.1016/j.asoc.2015.06.009 ↗
6. 10.1016/j.cogsys.2018.07.01510.1016/j.cogsys.2018.07.015 ↗
7. Liu X. Li B. Su Q. et al. (2007). A new accurate labeling algorithm for connected regions of binary images[J]. Computer Engineering and Applications 42(22) 76–78. (in Chinese with English abstract).
8. 10.1177/155892501986612110.1177/1558925019866121 ↗
9. 10.1177/004051751772302710.1177/0040517517723027 ↗
10. Ma, C. X., Liu, X. N., & Liu, F. (2014). Research on the statistical analysis of expert identification characteristics of wool fiber and automatic identification method. Wool Textile Science and Technology, 42(10), 62–64.
11. 10.1109/79.9121710.1109/79.91217 ↗
12. 10.1109/TSMC.1979.431007610.1109/TSMC.1979.4310076 ↗
13. 10.13475/j.fzxb.2008.01.013.(in10.13475/j.fzxb.2008.01.013.(in ↗
14. 10.1088/1742-6596/1948/1/01201010.1088/1742-6596/1948/1/012010 ↗
15. 10.1016/j.eswa.2023.12154910.1016/j.eswa.2023.121549 ↗
16. 10.1016/j.apacoust.2021.10857210.1016/j.apacoust.2021.108572 ↗
17. 10.14162/j.cnki.11-4772/t.2017.01.02010.14162/j.cnki.11-4772/t.2017.01.020 ↗
18. 10.19507/j.cnki.1673-0356.2020.09.003.(in10.19507/j.cnki.1673-0356.2020.09.003.(in ↗
19. 10.1117/12.300145310.1117/12.3001453 ↗
20. 10.1016/j.micron.2019.01.00910.1016/j.micron.2019.01.009 ↗
21. 10.1016/j.measurement.2019.04.01510.1016/j.measurement.2019.04.015 ↗
22. 10.19333/j.mfkj.2022040090910.19333/j.mfkj.20220400909 ↗
23. 10.4028/www.scientific.net/AMR.332-334.119810.4028/www.scientific.net/AMR.332-334.1198 ↗
24. 10.1016/j.precisioneng.2022.05.00610.1016/j.precisioneng.2022.05.006 ↗
25. 10.4028/www.scientific.net/KEM.671.38510.4028/www.scientific.net/KEM.671.385 ↗
26. 10.1080/00405000.2021.191439910.1080/00405000.2021.1914399 ↗
27. 10.1016/j.neucom.2024.12732310.1016/j.neucom.2024.127323 ↗
28. 10.1080/15440478.2023.217510010.1080/15440478.2023.2175100 ↗
29. 10.13509/j.cnki.ib.2012.02.00110.13509/j.cnki.ib.2012.02.001 ↗
30. 10.1016/j.heliyon.2024.e3453710.1016/j.heliyon.2024.e34537 ↗
31. 10.1177/0040517521104606010.1177/00405175211046060 ↗
32. 10.1177/1558925021100933310.1177/15589250211009333 ↗
33. 10.1177/1558925022114654810.1177/15589250221146548 ↗