Geographical Origin Determination of Ethiopian Teff Using Multi-Channel Capsule Networks
DOI:
https://doi.org/10.20372/pjet.v4i2.3700Keywords:
Multi-Channel, Capsule Network, Teff, Computer visionAbstract
Agriculture plays a vital role in food security and economic development in developing countries like Ethiopia. Precisely identifying the geographical origin of agricultural products like Teff is challenging because different varieties may have the same size, color, and shape. This study applies computer vision with multi-channel capsule network algorithm architecture to identify the geographical origin of Ethiopian Teff. The proposed approach overcomes the traditional CNN's challenges related to viewpoint variation, overfitting, and spatial variation for red and mixed varieties of Teff. The images of Teff were collected from various locations in the Amhara region of Ethiopia. Image preprocessing techniques used in this work include median, Gaussian, and non-local means filtering to enhance the image quality. Content-aware resizing was applied to resize the images, and Contrast Limited Adaptive Histogram Equalization (CLAHE) was used to enhance image contrast. For image segmentation, we used multi-Otsu thresholding, region-based segmentation, watershed segmentation, and U-Net to identify the regions of interest in the Teff images. Subsequently, we extracted features from the segmented images using the BRISK and HOG methods. Once the feature vectors were obtained, we developed a model using Capsule Networks to determine the geographical origin of Ethiopian Teff. A total of 11 distinct experiments took place and achieved notable results in three different approaches. The first approach, using the BRISK feature extraction method, combined Multi-Otsu thresholding, region-based segmentation, watershed segmentation, and comprehensive noise removal, achieved the highest accuracy of 84%. In the second approach, when combining region-based segmentation with that of median and Gaussian filtering, along with BRISK feature extraction, achieved an accuracy of 80%. The third approach, using Histogram of Oriented Gradients (HOG) feature extraction, combined with Multi-Otsu thresholding, region-based segmentation, watershed segmentation, and comprehensive noise removal, achieved a good accuracy of 92.5%. These findings demonstrate that our proposed model can accurately identify the geographical origin of Ethiopian Teff with a test accuracy of 92.5% performance.
References
1. Ali UM, Rahimi M. Evolution and Transformation of Ethiopian agriculture: A review of historical trends, contemporary challenges, and future opportunities. Turkish Journal of Agriculture - Food Science and Technology [Internet]. 2025 Dec 16;13(s3):3995–4011. Available from: https://doi.org/10.24925/turjaf.v13is3.3995-4011.8100
2. Melese D, Asfaw Z, Woldu Z, Warkineh B, Amare E, McAlvay AC, et al. Mixed teff (Eragrostis tef, Poaceae) cultivation and consumption among smallholder farmers in South Wollo Zone, Ethiopia. Journal of Ethnobiology and Ethnomedicine [Internet]. 2025 Apr 24;21(1):27. Available from: https://doi.org/10.1186/s13002-025-00776-2
3. Menjaye TK, Assefa K, Legese O, Seifu M, Kebede W, Abebe W, et al. Grain and flour physical and functional characteristics of Tef [Eragrostis tef (Zucc) Trotter] varieties of Ethiopia. Journal of Cereal Science [Internet]. 2024 Sep 23;120:104030. Available from: https://doi.org/10.1016/j.jcs.2024.104030
4. Reta C, Atlabachew M, Mehari B, Hilawea KT, Asmellash T. Discrimination of the geographical origin of gluten-free teff grains from northwestern parts of Ethiopia by fatty acid analysis. Heliyon [Internet]. 2024 Jan 22;10(3):e24932. Available from: https://doi.org/10.1016/j.heliyon.2024.e24932
5. Albahar M. A survey on deep learning and its impact on agriculture: Challenges and opportunities. Agriculture [Internet]. 2023 Feb 23;13(3):540. Available from: https://doi.org/10.3390/agriculture13030540
6. Cao Z, Sun S, Bao X. A review of computer vision and deep learning applications in crop growth management. Applied Sciences [Internet]. 2025 Jul 30;15(15):8438. Available from: https://doi.org/10.3390/app15158438
7. Arifah FN, Kusrini, Utami E, Wibowo FW, Chandramouli K. Vision-Based deep learning for crop classification in precision agriculture [Internet]. IEEE; 2025. Available from: https://doi.org/10.1109/iccit65724.2025.11167352
8. Chuquimarca LE, Vintimilla BX, Velastin SA. A review of external quality inspection for fruit grading using CNN models. Artificial Intelligence in Agriculture [Internet]. 2024 Oct 16;14:1–20. Available from: https://doi.org/10.1016/j.aiia.2024.10.002
9. Dönmez E, Diker A, Elen A, Ulu M. Multiple deep learning by majority-vote to classify haploid and diploid maize seeds. Scientia Horticulturae [Internet]. 2024 Aug 27;337:113549. Available from: https://doi.org/10.1016/j.scienta.2024.113549
10. Math RM, Dharwadkar NV, Sajjanar V, Deshmukh MN. Deep Learning for Precision Grade-Quality Classification of Pomegranates [Internet]. IEEE; 2024. Available from: https://doi.org/10.1109/innova63080.2024.10846982
11. Khan A, Hassan T, Shafay M, Fahmy I, Werghi N, Mudigansalage S, et al. Tomato maturity recognition with convolutional transformers. Scientific Reports [Internet]. 2023 Dec 21;13(1):22885. Available from: https://doi.org/10.1038/s41598-023-50129-w
12. Woldetsadik D, Sims DB, Garner MC, Hailu H. Metal(loid)s Profile of Four Traditional Ethiopian Teff Brands: Geographic Origin Discrimination. Biological Trace Element Research [Internet]. 2023 Jun 28;202(3):1305–15. Available from: https://doi.org/10.1007/s12011-023-03736-7
13. Ermyas A, Debasu. Developing an automatic ethiopian teff identification system using machine learning approach [Internet]. 2020. Available from: http://ir.bdu.edu.et/handle/123456789/12724
14. Bhatt D, Patel C, Talsania H, Patel J, Vaghela R, Pandya S, et al. CNN variants for Computer Vision: history, architecture, application, challenges and future scope. Electronics [Internet]. 2021 Oct 11;10(20):2470. Available from: https://doi.org/10.3390/electronics10202470
15. Patrick MK, Adekoya AF, Mighty AA, Edward BY. Capsule Networks – A survey. Journal of King Saud University - Computer and Information Sciences [Internet]. 2019 Sep 25;34(1):1295–310. Available from: https://doi.org/10.1016/j.jksuci.2019.09.014
16. Alajrami MA, Abu-Naser SS. Type of tomato classification using deep learning. PhilPapers (PhilPapers Foundation) [Internet]. 2020 Jan 1;3(12). Available from: https://philarchive.org/rec/ALATOT
17. Gope HL, Fukai H. Normal and Peaberry Coffee Beans Classification from Green Coffee Bean Images Using Convolutional Neural Networks and Support Vector Machine. Zenodo (CERN European Organization for Nuclear Research) [Internet]. 2020 Jul 14;14(6):189–96. Available from: https://zenodo.org/record/3943675
18. Bedaso M, Meshesha M, Diriba C. Grading Ethiopian coffee raw quality using image processing techniques. Research Square [Internet]. 2022 Aug 24; Available from: https://doi.org/10.21203/rs.3.rs-1980632/v1
19. Saranya N, Srinivasan K, Kumar SKP. Banana ripeness stage identification: a deep learning approach. Journal of Ambient Intelligence and Humanized Computing [Internet]. 2021 May 3;13(8):4033–9. Available from: https://doi.org/10.1007/s12652-021-03267-w
20. Addisalem T, Biazen. DEVELOPING AN AUTOMATIC SHIRO FLOUR VARIETY RECOGNITION MODEL USING a COMBINED FEATURES OF CNN AND GLCM [Internet]. 2022. Available from: http://ir.bdu.edu.et/handle/123456789/14381
21. Ermyas A, Debasu. DEVELOPING AN AUTOMATIC ETHIOPIAN TEFF IDENTIFICATION SYSTEM USING MACHINE LEARNING APPROACH [Internet]. 2020. Available from: http://ir.bdu.edu.et/handle/123456789/12724
22. Wang Q, Yuan Y. Learning to resize image. Neurocomputing [Internet]. 2013 Dec 11;131:357–67. Available from: https://doi.org/10.1016/j.neucom.2013.10.007
23. Danon D, Arar M, Cohen-Or D, Shamir A. Image resizing by reconstruction from deep features. Computational Visual Media [Internet]. 2021 Apr 27;7(4):453–66. Available from: https://doi.org/10.1007/s41095-021-0216-x
24. Abdelatif MH, Madani GA, Al-Arbash AK, Oshah AA. Making noise reduction in medical imaging more accessible: machine learning and traditional techniques. Journal of Technology Research [Internet]. 2024 Dec 15;2(2):75–80. Available from: https://doi.org/10.26629/jtr.2024.10
25. Ayubi J, Amirani MC, Valizadeh M. A new content-aware image resizing based on Rényi entropy and deep learning. Neural Computing and Applications [Internet]. 2024 Mar 30;36(15):8885–99. Available from: https://doi.org/10.1007/s00521-024-09517-0
26. Aijaz O, Ali SM, Uyghur Y. Analysis of different algorithmic design techniques for seam carving. arXiv (Cornell University) [Internet]. 2024 Oct 28; Available from: http://arxiv.org/abs/2410.21207
27. Khudhair ZN, Khdiar AN, Abbadi NKE, Mohamed F, Saba T, Alamri FS, et al. Color to grayscale image conversion based on singular value decomposition. IEEE Access [Internet]. 2023 Jan 1;11:54629–38. Available from: https://doi.org/10.1109/access.2023.3279734
28. Bataineh B. Brightness and contrast enhancement method for color images via pairing adaptive gamma correction and histogram equalization. International Journal of Advanced Computer Science and Applications [Internet]. 2023 Jan 1;14(3). Available from: https://doi.org/10.14569/ijacsa.2023.0140314
29. Jing Z, Tang B. Improved image segmentation method based on Otsu thresholding and level set techniques. Journal of Physics Conference Series [Internet]. 2024 Aug 1;2813(1):012017. Available from: https://doi.org/10.1088/1742-6596/2813/1/012017
30. Sang G, Wang X, Zhang J. Improved OTSU Theory of image multi-threshold segmentation by incorporating AnT Colony Algorithm. Informatica [Internet]. 2024 Jun 4;48(9). Available from: https://doi.org/10.31449/inf.v48i9.5872
31. Soomro S, Niaz A, Soomro TA, Kim J, Manzoor A, Choi KN. Selective image segmentation driven by region, edge and saliency functions. PLoS ONE [Internet]. 2023 Dec 15;18(12):e0294789. Available from: https://doi.org/10.1371/journal.pone.0294789
32. Ça?lan ?, De?irmenci A, Çankaya ?, Gökta? ÖF. Application of watershed segmentation on biomedical images as an educational tool. F?rat Üniversitesi Mühendislik Bilimleri Dergisi [Internet]. 2025 Sep 20;37(2):763–80. Available from: https://doi.org/10.35234/fumbd.1706496
33. Bhatti MA, Syam, Chen H, Hu Y, Keung LW, Zeeshan Z, et al. Utilizing convolutional neural networks (CNN) and U-Net architecture for precise crop and weed segmentation in agricultural imagery: A deep learning approach. Big Data Research [Internet]. 2024 May 1;36:100465. Available from: https://doi.org/10.1016/j.bdr.2024.100465
34. Zhang L, Xu W, Shen C, Huang Y. Vision-Based On-Road nighttime vehicle detection and tracking using improved HOG features. Sensors [Internet]. 2024 Feb 29;24(5):1590. Available from: https://doi.org/10.3390/s24051590
35. Matuzevi?ius D, Urbanavi?ius V, Miniotas D, Miku?ionis Š, Laptik R, Ušinskas A. Key-Point-Descriptor-Based image quality evaluation in photogrammetry workflows. Electronics [Internet]. 2024 May 29;13(11):2112. Available from: https://doi.org/10.3390/electronics13112112
36. Kumawat A, Panda S, Gerogiannis VC, Kanavos A, Acharya B, Manika S. A hybrid approach for image acquisition methods based on Feature-Based image registration. Journal of Imaging [Internet]. 2024 Sep 14;10(9):228. Available from: https://doi.org/10.3390/jimaging10090228
37. Li P, Fei Q, Chen Z, Liu X. Interpretable Multi-Channel Capsule Network for Human Motion Recognition. Electronics [Internet]. 2023 Oct 18;12(20):4313. Available from: https://doi.org/10.3390/electronics12204313
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Poly Journal of Engineering and Technology (PJET)
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
