Heading Classification of Rice Cultivar Groups Using Time-Series RGB Drone Images and a Deep-Learning-Based Classification Model

Yunjeong Jeong; Hyeonggyun Kim; Do-hyun Kim; Wangyu Sang; Dongwon Kwon; Woojin Im; Dasom Jeon; Huisu Bae; Sungyul Chang

doi:10.7740/kjcs.2026.71.3.098

All Issue

2026 Vol.71, Issue 3 Preview Page Next Page

Original Research Article

Heading Classification of Rice Cultivar Groups Using Time-Series RGB Drone Images and a Deep-Learning-Based Classification Model 드론 시계열 RGB 영상과 딥러닝 기반 분류 모델을 이용한 벼 품종군의 출수 판별

1 September 2026. pp. 98-105

PDF XML

Abstract

The rice heading date is a critical phenological indicator when optimizing crop management and determining harvest timing. Traditional assessments rely on manual visual inspections, which are prone to observer bias and are difficult to scale in breeding programs. In this study, we evaluated the feasibility of automated heading date estimation using unmanned aerial vehicle (UAV)-based time-series red-green-blue (RGB) imagery. This study was conducted using 144 rice cultivars at the National Institute of Crop Science, Wanju-gun, Jeollabuk-do, Republic of Korea. RGB imagery was acquired at three time points (08:00, 12:00, and 16:00), processed into orthomosaics, and partitioned into 100 × 100 pixel patches to create a dataset of 79,331 images. These images were labeled as pre-heading or heading, based on ground truth data. A YOLOv11-based classification model, initialized using ImageNet pre-trained weights, was employed, and its decision-making process was visualized using Grad-CAM. The model achieved a final accuracy of 0.953 and training loss of 0.142. Although the confusion matrix demonstrated high overall classification performance, misclassifications were primarily concentrated in the early heading stage. A comparison between predicted and observed heading dates across maturity groups indicated statistically significant differences. The predicted heading dates had a mean bias of 1.89 days, indicating a tendency to estimate heading dates later than the field observations. The mean absolute error and root-mean-square error were 2.14 and 2.66 days, respectively, and the predicted heading dates of 60.42% of the cultivars were within ±2 days of the observed heading dates. Grad-CAM analysis confirmed that the model focused on panicle features as the primary classification criterion. These findings suggest that UAV-based YOLO models have potential as a phenotyping support tool for consistent heading date assessment across large numbers of rice cultivars. Further refinement through additional early-heading data and improved temporal decision criteria may decrease prediction delays and increase the applicability of the model to automated crop growth monitoring.

Keywords

deep learning

drone imagery

grad-CAM

rice heading date

YOLO

References

Bleiholder, H., E. Weber, P. Lancashire, C. Feller, L. Buhr, M. Hess, H. Wicke, H. Hack, U. Meier, and R. Klose. 2001. Growth stages of mono-and dicotyledonous plants. BBCH monograph 158 : 20180906-20074619.

Chaurasia, H., A. Arora, D. Raju, S. Marwaha, V. Chinnusamy, R. Jain, and M. A. Haque. 2024. PanicleDet: a deep learning-based model for detection of panicle stages in paddy. Neural Comput. Appl. 37 : 2663-2673.

10.1007/s00521-024-10746-6

Chen, R., H. Lu, Y. Wang, Q. Tian, C. Zhou, A. Wang, Q. Feng, S. Gong, Q. Zhao, and B. Han. 2024. High-throughput UAV-based rice panicle detection and genetic mapping of heading-date-related traits. Front. Plant Sci. 15 : 1327507.

10.3389/fpls.2024.132750738562563PMC10984267

Counce, P. A., T. C. Keisling, and A. J. Mitchell. 2000. A Uniform, Objective, and Adaptive System for Expressing Rice Development. Crop Sci. 40 : 436-443.

10.2135/cropsci2000.402436x

Desai, S. V., V. N. Balasubramanian, T. Fukatsu, S. Ninomiya, and W. Guo. 2019. Automatic estimation of heading date of paddy rice using deep learning. Plant Methods 15 : 76.

10.1186/s13007-019-0457-131338116PMC6626381

Duan, L., C. Huang, G. Chen, L. Xiong, Q. Liu, and W. Yang. 2015. Determination of rice panicle numbers during heading by multi-angle imaging. Crop J. 3 : 211-219.

10.1016/j.cj.2015.03.002

Guo, W., T. Fukatsu, and S. Ninomiya. 2015. Automated characterization of flowering dynamics in rice using field-acquired time-series RGB images. Plant Methods 11 : 7.

10.1186/s13007-015-0047-925705245PMC4336727

Guo, Y., W. Zhan, Z. Zhang, Y. Zhang, and H. Guo. 2025. FRPNet: A Lightweight Multi-Altitude Field Rice Panicle Detection and Counting Network Based on Unmanned Aerial Vehicle Images. Agronomy 15 : 1396.

10.3390/agronomy15061396

Hoshikawa, K. 1989. The growing rice plant : an anatomical monograph. Nobunkyo Tokyo.

Kim, J. H., S. H. Jo, J. H. Moon, S. Y. Yang, J. K. Baek, Y. S. Song, J. Y. Shon, and H. S. Lee. 2025. New insights into rice phenology: discovering the effect of insolation on heading response. Physiol. Plant. 177 : e70132.

10.1111/ppl.7013239968699PMC11836917

Lecun, Y., Y. Bengio, and G. Hinton. 2015. Deep learning. Nature 521 : 436-444.

10.1038/nature14539

Lee, C.-M., Y.-H. Kwon, H.-M. Park, J.-U. Jeung, H.-S. Park, M.-K. Baek, S.-K. Ha, and Y. Mo. 2020. Days to Heading and Culm Length Variation of Korean Rice Varieties in Different Environments. Korean J. Breed. Sci. 52 : 389-397.

10.9787/KJBS.2020.52.4.389

Li, S., Y. Luo, G. Wei, W. Zong, W. Zeng, D. Xiao, H. Zhang, Y. Song, Y. Hao, and K. Sun. 2023. Improving yield-related traits by editing the promoter of the heading date gene Ehd1 in rice. Theor. Appl. Genet. 136: 239.

10.1007/s00122-023-04489-6

Lu, X., Y. Shen, J. Xie, X. Yang, Q. Shu, S. Chen, Z. Shen, and H. Cen. 2024. Phenotyping of panicle number and shape in rice breeding materials based on unmanned aerial vehicle imagery. Plant Phenomics 6 : 0265.

10.34133/plantphenomics.026539449974PMC11499587

Luo, Y., X. Li, B. Bai, X. Yu, Y. Wang, Z. Ma, L. Zhang, and X. Peng. 2025. A field rice panicle detection model based on improved YOLOv11x. Front. Plant Sci. 16 : 1656505.

10.3389/fpls.2025.165650540963833PMC12436299

Lyu, M., X. Lu, Y. Shen, Y. Tan, L. Wan, Q. Shu, Y. He, Y. He, and H. Cen. 2023. UAV time-series imagery with novel machine learning to estimate heading dates of rice accessions for breeding. Agric. For. Meteorol. 341 : 109646.

10.1016/j.agrformet.2023.109646

Redmon, J., S. Divvala, R. Girshick, and A. Farhadi. 2016. You only look once: Unified, real-time object detection. IEEE conference on computer vision and pattern recognition. pp. 779-788.

10.1109/CVPR.2016.91

Selvaraju, R. R., M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra. 2017. Grad-cam: Visual explanations from deep networks via gradient-based localization. IEEE international conference on computer vision. pp. 618-626.

10.1109/ICCV.2017.74

Sun, B., Y. Li, J. Huang, Z. Cao, and X. Peng. 2024. Impacts of variable illumination and image background on rice LAI estimation based on UAV RGB-derived color indices. Appl. Sci. 14 : 3214.

10.3390/app14083214

Sun, B., W. Zhou, S. Zhu, S. Huang, X. Yu, Z. Wu, X. Lei, D. Yin, H. Xia, and Y. Chen. 2022. Universal detection of curved rice panicles in complex environments using aerial images and improved YOLOv4 model. Front. Plant Sci. 13 : 1021398.

10.3389/fpls.2022.102139836420030PMC9676644

Tsouros, D. C., S. Bibi, and P. G. Sarigiannidis. 2019. A review on UAV-based applications for precision agriculture. Information. 10 : 349.

10.3390/info10110349

Wang, X., W. Yang, Q. Lv, C. Huang, X. Liang, G. Chen, L. Xiong, and L. Duan. 2022. Field rice panicle detection and counting based on deep learning. Front. Plant Sci. 13 : 966495.

10.3389/fpls.2022.96649536035660PMC9416702

Wu, H., M. Guan, J. Chen, Y. Pan, J. Zheng, Z. Jin, H. Li, and S. Tan. 2025. OE-YOLO: An EfficientNet-Based YOLO Network for Rice Panicle Detection. Plants 14 : 1370.

10.3390/plants1409137040364399PMC12074215

Xiong, X., L. Duan, L. Liu, H. Tu, P. Yang, D. Wu, G. Chen, L. Xiong, W. Yang, and Q. Liu. 2017. Panicle-SEG: a robust image segmentation method for rice panicles in the field based on deep learning and superpixel optimization. Plant Methods 13 : 104.

10.1186/s13007-017-0254-729209408PMC5704426

Ye, T., S. Zong, A. Kleidon, W. Yuan, Y. Wang, and P. Shi. 2019. Impacts of climate warming, cultivar shifts, and phenological dates on rice growth period length in China after correction for seasonal shift effects. Clim. Change. 155 : 127-143.

10.1007/s10584-019-02450-5

Zhou, C., W. Wang, H. Zheng, M. Huang, and J. Chen. 2025. Effects of harvest time on rice yield and quality: A meta-analysis. Eur. J. Agron. 169 : 127674.

10.1016/j.eja.2025.127674

Zhou, S., S. Zhu, S. Cui, H. Hou, H. Wu, B. Hao, L. Cai, Z. Xu, L. Liu, and L. Jiang. 2021. Transcriptional and post‐transcriptional regulation of heading date in rice. New Phytol. 230 : 943-956.

10.1111/nph.1715833341945PMC8048436

Information

Publisher :The Korean Society of Crop Science
Publisher(Ko) :한국작물학회
Journal Title :The Korean Journal of Crop Science
Journal Title(Ko) :한국작물학회지
Volume : 71
No :3
Pages :98-105
Received Date : 2026-06-17
Revised Date : 2026-07-17
Accepted Date : 2026-07-23
DOI :https://doi.org/10.7740/kjcs.2026.71.3.098

The Korean Journal of Crop Science ISSN:0252-9777(Print) 2287-8432(Online) 한국작물학회지

All Issue