All Issue

2025 Vol.70, Issue 4 Preview Page

Original Research Article

Comparison of Deep Learning and Vegetation Index Methodologies for Estimating Rice Canopy Coverage Based on RGB Images

RGB 영상 기반 시계열 벼 군락 피복도 추정을 위한 딥러닝 모델과 식생지수 방법론 비교

Eun-Ji Kim1
,
Do-Hyun Kim2
,
In-Ha Lee1
,
Sungyul Chang3
,
Dongwon Kwon3
,
Woo-jin Im3
,
Woon-Ha Hwang4
,
Wan-Gyu Sang3
,
Jae-Ki Chang4
,
Hyeok-jin Bak5*
김 은지1
,
김 도현2
,
이 인하1
,
장 성율3
,
권 동원3
,
임 우진3
,
황 운하4
,
상 완규3
,
장 재기4
,
박 혁진5*
1Student, National Institute of Crop and Food Science, Rural Development Administration, RDA, Wanju 55365, Korea
2Private-sector Expert, National Institute of Crop and Food Science, Rural Development Administration, RDA, Wanju 55365, Korea
3Junior Scientist, National Institute of Crop and Food Science, Rural Development Administration, RDA, Wanju 55365, Korea
4Senior Scientist, National Institute of Crop and Food Science, Rural Development Administration, RDA, Wanju 55365, Korea
5Professional Researcher, National Institute of Crop and Food Science, Rural Development Administration, RDA, Wanju 55365, Korea
1국립식량과학원 산학연학생
2국립식량과학원 민간전문가
3국립식량과학원 농업연구사
4국립식량과학원 농업연구관
5국립식량과학원 전문연구원
*Corresponding Author
1 December 2025. pp. 202-212
PDF XML Citation
Abstract

Canopy coverage is a key agronomic trait for assessing rice growth and predicting grain yield. However, conventional measurement methods are labor-intensive and inefficient. This study develop a robust, automated system for estimating rice canopy coverage by comparing traditional image analysis techniques with deep learning-based semantic segmentation models. The methodologies compared were: (1) five traditional vegetation index-based methods and (2) five mainstream convolutional neural network (CNN)-based semantic segmentation models (i.e., U-net, DeepLabV3+, LinkNet, FPN, PSPNet), each implemented with a ResNet-101 backbone. A large-scale dataset of 5,000 RGB images captured from 18 locations across Korea was constructed for training and validation. Deep learning-based models significantly outperformed the vegetation index methods. Among the CNN models, U-net achieved the highest performance with a mean Intersection over Union (mIOU) of 0.986. This study identifies U-net architecture as an optimal methodology for automated and highly accurate canopy coverage analysis and establishes a foundational technology for advancing high-throughput phenotyping in precision agriculture.

Keywords
canopy coverage
deep learning
rice
semantic segmentation
CNN
References
1

Bak, H. J., D. W. Kwon, W. G. Sang, H. Y. Ban, S. Y. Chang, J. K. Beak, Y. H. Lee, W. J. Im, M. C. Seo, and J. I. Cho. 2023. Deep learning approaches for accurate weed Area assessment in maize fields. Korean J. Agric. For. Meteorol. 25(1) : 17-27. https://doi.org/10.5532/KJAFM.2023.25.1.17

10.5532/KJAFM.2023.25.1.17
2

Bendig, J., K. Yu, H. Aasen, A. Bolten, S. Bennertz, J. Broscheit, M.L. Gnyp, and G. Bareth. 2015. Combining UAV-based plant height from crop surface models, visible, and near infrared vegetation indices for biomass monitoring in barley. Int. J. Appl. Earth Obs. Geoinf. 39 : 79-87. https://doi.org/10.1016/j.jag.2015.02.012

10.1016/j.jag.2015.02.012
3

Bréda, N. J. J. 2003. Ground-based measurements of leaf area index: A review of methods, instruments and current controversies. J. Exp. Bot. 54(392) : 2403-2417. https://doi.org/10.1093/jxb/erg263

10.1093/jxb/erg263
4

Chaurasia, A., and E. Culurciello. 2017. LinkNet: Exploiting encoder representations for efficient 614 semantic segmentation. arXiv preprint arXiv:1707.03718. https://arxiv.org/abs/1707.03718

10.1109/VCIP.2017.8305148
5

Chen, L.C., G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille. 2018. DeepLab: Semantic 616 image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. 617 IEEE Trans. Pattern Anal. Mach. Intell. 40(4) : 834-848. https://doi.org/10.1109/TPAMI.2017.2699184

10.1109/TPAMI.2017.2699184
6

Das Choudhury, S., A. Samal, and T. Awada. 2019. Leveraging image analysis for high-throughput plant phenotyping. Front. Plant Sci. 10: 508.

10.3389/fpls.2019.0050831068958PMC6491831
7

Ge, H., G. Lv, Y. Qin, and M. Shen. 2025. Assessing leaf nitrogen concentration in rice using RGB imaging: a comparative study at leaf, canopy, and plot scales. Front. Plant Sci. 16 : 1599177.

10.3389/fpls.2025.159917740838074PMC12361148
8

Kim, E. J., H. J. Bak, D. Kwon, S. Chang, W. J. Im, N. J. Chung, J. K. Chang, W. H. Hwang, and W. G. Sang. 2025. Rice leaf area index evaluation using rgb-based vegetation indices and machine learning models. Korean J. Crop Sci. 70(2) : 57-67. https://doi.org/10.7740/kjcs.2025.70.2.057

10.7740/kjcs.2025.70.2.057
9

Lee, K. J. and B. W. Lee. 2011. Estimating canopy cover from color digital camera image of rice field. J. Crop Sci. Biotechnol. 14(2) : 151-155. https://doi.org/10.1007/s12892-011-0029-z

10.1007/s12892-011-0029-z
10

Liao, J., Y. Wang, J. Yin, L. Liu, S. Zhang, and D. Zhu. 2018. Segmentation of rice seedlings using the YCrCb color space and an improved otsu method. Agronomy 8(11) : 269. https:// doi.org/10.3390/agronomy8110269

10.3390/agronomy8110269
11

Lin, T. Y., P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie. 2017. Feature pyramid 661 networks for object detection. arXiv preprint arXiv:1612.03144. https://arxiv.org/abs/1612.03144

12

Louhaichi, M., M. M. Borman, and D. E. Johnson. 2001. Spatially located platform and aerial photography for documentation of grazing impacts on wheat. Geocarto Int. 16(1) : 65-70. https:// doi.org/10.1080/10106040108542184

10.1080/10106040108542184
13

Ma, X., X. Deng, L. Qi, Y. Jiang, H. Li, Y. Wang, and X. Xing. 2019. Fully convolutional network for rice seedling and weed image segmentation at the seedling stage in paddy fields. PloS one 14(4) : e0215676. https://doi.org/10.1371/journal.pone.0215676

10.1371/journal.pone.021567630998770PMC6472823
14

Otsu, N. 1979. A threshold selection method from gray-level histograms. IEEE Transactions on Systems, Man, and Cybernetics, 9(1) : 62-66. https://doi.org/10.1109/TSMC.1979.4310076

10.1109/TSMC.1979.4310076
15

Reza, M. N., I. S. Na, S. W. Baek, and K. H. Lee. 2019. Rice yield estimation based on K-means clustering with graph-cut segmentation using low-altitude UAV images. Biosystems Engineering, 177, 109-122. https://doi.org/10.1016/j.biosystemseng.2018.09.014

10.1016/j.biosystemseng.2018.09.014
16

Ronneberger, O., P. Fischer, and T. Brox. 2015. U-Net: Convolutional networks for biomedical 681 image segmentation. arXiv preprint arXiv:1505.04597. https://arxiv.org/abs/1505.04597

17

Tucker, C. J. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8(2) : 127-150. https://doi.org/10.1016/0034-4257(79)90013-0

10.1016/0034-4257(79)90013-0
18

Woebbecke, D. M., G. E. Meyer, K. Von Bargen, and D. A. Mortensen. 1995. Color indices for weed identification under various soil, residue, and lighting conditions. Transactions of the ASAE 38(1) : 259-269. https://doi.org/10.13031/2013.27838

10.13031/2013.27838
19

Zhang, Y., Y. Jiang, B. Xu, G. Yang, H. Feng, X. Yang, H. Yang, C. Liu, Z. Cheng, and Z. Feng. 2024. Study on the estimation of leaf area index in rice based on UAV RGB and multispectral data. Remote Sens. 16(16) : 3049.

10.3390/rs16163049
20

Zhao, H., J. Shi, X. Qi, X. Wang, and J. Jia. 2017. Pyramid scene parsing network. arXiv preprint 720 arXiv:1612.01105. https://arxiv.org/abs/1612.01105

Information
  • Publisher :The Korean Society of Crop Science
  • Publisher(Ko) :한국작물학회
  • Journal Title :The Korean Journal of Crop Science
  • Journal Title(Ko) :한국작물학회지
  • Volume : 70
  • No :4
  • Pages :202-212
  • Received Date : 2025-10-22
  • Revised Date : 2025-11-16
  • Accepted Date : 2025-11-18
  • DOI :https://doi.org/10.7740/kjcs.2025.70.4.202
Journal Informaiton The Korean Journal of Crop Science The Korean Journal of Crop Science
Online Submission
  • NRF
  • KOFST
  • KISTI Current Status
  • KISTI Cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close