DOI QR코드

DOI QR Code

Deep Learning Approaches for Accurate Weed Area Assessment in Maize Fields

딥러닝 기반 옥수수 포장의 잡초 면적 평가

  • Hyeok-jin Bak (National Institute of Crop Science, Rural Development Administration) ;
  • Dongwon Kwon (National Institute of Crop Science, Rural Development Administration) ;
  • Wan-Gyu Sang (National Institute of Crop Science, Rural Development Administration) ;
  • Ho-young Ban (National Institute of Crop Science, Rural Development Administration) ;
  • Sungyul Chang (National Institute of Crop Science, Rural Development Administration) ;
  • Jae-Kyeong Baek (National Institute of Crop Science, Rural Development Administration) ;
  • Yun-Ho Lee (National Institute of Crop Science, Rural Development Administration) ;
  • Woo-jin Im (National Institute of Crop Science, Rural Development Administration) ;
  • Myung-chul Seo (National Institute of Crop Science, Rural Development Administration) ;
  • Jung-Il Cho
  • 박혁진 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 권동원 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 상완규 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 반호영 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 장성율 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 백재경 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 이윤호 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 임우진 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 서명철 (농촌진흥청 국립식량과학원 작물재배생리과) ;
  • 조정일 (농촌진흥청 국립식량과학원 작물재배생리과)
  • Received : 2023.02.16
  • Accepted : 2023.03.27
  • Published : 2023.03.30

Abstract

Weeds are one of the factors that reduce crop yield through nutrient and photosynthetic competition. Quantification of weed density are an important part of making accurate decisions for precision weeding. In this study, we tried to quantify the density of weeds in images of maize fields taken by unmanned aerial vehicle (UAV). UAV image data collection took place in maize fields from May 17 to June 4, 2021, when maize was in its early growth stage. UAV images were labeled with pixels from maize and those without and the cropped to be used as the input data of the semantic segmentation network for the maize detection model. We trained a model to separate maize from background using the deep learning segmentation networks DeepLabV3+, U-Net, Linknet, and FPN. All four models showed pixel accuracy of 0.97, and the mIOU score was 0.76 and 0.74 in DeepLabV3+ and U-Net, higher than 0.69 for Linknet and FPN. Weed density was calculated as the difference between the green area classified as ExGR (Excess green-Excess red) and the maize area predicted by the model. Each image evaluated for weed density was recombined to quantify and visualize the distribution and density of weeds in a wide range of maize fields. We propose a method to quantify weed density for accurate weeding by effectively separating weeds, maize, and background from UAV images of maize fields.

포장에서 잡초의 발생은 농작물의 생산량을 크게 떨어트리는 원인 중 하나이고 SSWM을 기반으로 잡초를 변량 방제하기 위해서 잡초의 발생 위치, 밀도 그리고 이를 정량화하는 것은 필수적이다. 본 연구에서는 2020년의 국립식량과학원에서 잡초 피해를 입은 옥수수 포장의 영상데이터를 무인항공기를 활용해서 수집하였고 이를 배경과 옥수수로 분리하여 딥러닝 기반 영상 분할 모델 제작을 위한 학습데이터를 획득하였다. DeepLabV3+, U-Net, Linknet, FPN의 4가지의 영상 분할 네트워크들의 옥수수의 검출 정확도를 평가하기 위해 픽셀정확도, mIOU, 정밀도, 재현성의 지표를 활용해서 정확도를 검증하였다. 검증 결과 DeepLabV3+ 모델이 0.76으로 가장 높은 mIOU를 나타냈고, 해당 모델과 식물체의 녹색 영역과 배경을 분리하는 지수인 ExGR을 활용해서 잡초의 면적을 정량화, 시각화하였다. 이러한 연구의 결과는 무인항공기로 촬영된 영상을 활용해서 넓은 면적의 옥수수 포장에서 빠르게 잡초의 위치와 밀도를 특정하고 정량화하는 것으로 잡초의 밀도에 따른 제초제의 변량 방제를 위한 의사결정에 도움이 될 것으로 기대한다.

Keywords

Acknowledgement

본 논문은 농촌진흥청 공동연구사업(과제번호: PJ0151012022)의 지원에 의해 이루어진 것임.

References

  1. Choi, B. S., C. G. Kim, K. Y. Seong, D. Y. Song, W. T. Jeon, J. S. Cho, K. H. Jeong, and U. G. Kang, 2011: Change of Weed Community in No-till Corn with Legume Cover Crops as Living Mulch. Korean Journal of Weed Science 31, 34-40. (in Korean with English abstract) https://doi.org/10.5660/KJWS.2011.31.1.034
  2. Chen, L. C., Y. Zhu, G. Papandreou, F. Schroff, and H. Adam, 2018: Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation. Computer Vision and Pattern Recognition, arXiv:1802.02611
  3. Chaurasia, A., and E. Culurciello, 2017: LinkNet: Exploiting Encoder Representations for Efficient Semantic Segmentation. Computer Vision and Pattern Recognition, arXiv:1707.03718
  4. Christensen, S., H. T. Sogaard, P. Kudsk, M. Norremark, I. Lund, E. S. Nadimi, and R. Jorgensen, 2009: Site-specific weed control technologies. Weed Research 49, 233-241. https://doi.org/10.1111/j.1365-3180.2009.00696.x
  5. Du, M., and N. Noguchi, 2017: Monitoring of wheat growth status and mapping of wheat yield's within-field spatial variations using color images acquired from UAV-camera system. Remote Sensing 9, 289.
  6. Ge, L., Z. Yang, Z. Sun, G. Zhang, M. Zhang, K. Zhang, C. Zhang, Y. Tan, and W. Li, 2019: A method for broccoli seedling recognition in natural environment based on binocular stereo vision and Gaussian mixture model. Sensors 19(5), 1132.
  7. He, K., X. Zhang, S. Ren, and J. Sun, 2016: Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 770-778.
  8. Huang, H., J. Deng, Y. Lan, A. Yang, X. Deng, S. Wen, H. Zhang, and Y. Zhang, 2018: Accurate weed mapping and prescription map generation based on fully convolutional networks using UAV imagery. Sensors 18(10), 3299.
  9. Lee, B. M., J. R. Jo, N. H. An, J. H. Ok, and S. C. Kim, 2014: False Seedbed Weed Control under Different Preparation Date and Method in Organic Corn Field. The Korean Society of Weed Science and The Turfgrass Society of Korea 3(4), 299-304. (in Korean with English abstract) https://doi.org/10.5660/WTS.2014.3.4.299
  10. Kazmi, W., S. Foix, G. Alenya, and H. J. Andersen, 2014: Indoor and Outdoor Depth Imaging of Leaves with Time-of-Flight and Stereo Vision Sensors: Analysis and Comparison. ISPRS Journal of Photogrammetry and Remote Sensing 88, 128-146. https://doi.org/10.1016/j.isprsjprs.2013.11.012
  11. Khan, M. J., H. S. Khan, A. Yousaf, K. Khurshid, and A. Abbas, 2018: Modern Trends in Hyperspectral Image Analysis: A Review. IEEE Access 6, 14118-14129. https://doi.org/10.1109/ACCESS.2018.2812999
  12. Lammie, C., A. Olsen, T. Carrick, and M. R. Azghadi, 2019: Low-power and High-Speed Deep FPGA Inference Engines for Weed Classification at the Edge. IEEE Access 7, 51171-51184. https://doi.org/10.1109/ACCESS.2019.2911709
  13. Richard, G. G., 2003: Injury to Vegetable Crops from Herbicides Applied in Previous Years. Weed Technology 17, 73-78. https://doi.org/10.1614/0890-037X(2003)017[0073:ITVCFH]2.0.CO;2
  14. Lee, J., M. N. Shin, B. I. Ku, K. B. Shim, and W. T. Jeon, 2021: Current Status and Direction of Weed Management According to Cropping Systems. Korean Journal of Crop Science 66(4), 459-466. (in Korean with English abstract) https://doi.org/10.7740/KJCS.2021.66.4.459
  15. Lee, Y. H., W. G. Sang, J. K. Baek, J. H. Kim, J. I. Cho, and M. C. Seo, 2020: Low-cost Assessment of Canopy Light Interception and Leaf Area in Soybean Canopy Cover using RGB Color Images. Korean Journal of Agricultural and Forest Meteorology 22(1), 13-19. (in Korean with English abstract)
  16. Lee, Y. H., M. C. Seo, and J. K. Beak, 2019: Growth Characteristics and Grain Yield Under Tropical Monsoon Climate in Forage Maize. Journal of the Korean Society for International Agriculture 31(3), 255-260. (in Korean with English abstract) https://doi.org/10.12719/KSIA.2019.31.3.255
  17. Lopez-Granados, F., J. Torres-Sanchez, A. Serrano-Perez, A. I. de Castro, F. J. Mesas-Carrascosa, and J. M. Pena, 2016: Early Season Weed Mapping in Sunflower using UAV Technology: Variability of Herbicide Treatment Maps Against Weed Thresholds. Precision Agriculture 17, 183-199. https://doi.org/10.1007/s11119-015-9415-8
  18. Louargant, M., G. Jones, R. Faroux, J. N. Paoli, T. Maillot, C. Gee, and S. Villette, 2018: Unsupervised Classification Algorithm for Early Weed Detection in Row-Crops by Combining Spatial and Spectral Information. Remote Sensing 10(5), 761.
  19. Luis, P., and W. Jason, 2017: The Effectiveness of Data Augmentation in Image Classification using Deep Learning. Computer Vision and Pattern Recognition, arXiv:1712.04621
  20. Mateen, A., and Q. Zhu, 2019: Weed Detection in Wheat Crop using UAV for Precision Agriculture. Pakistan Journal of Agricultural Sciences 56(3), 775-784. https://doi.org/10.21162/PAKJAS/19.8036
  21. Nevavuori, P., N. Narra, and T. Lipping, 2019: Crop Yield Prediction with Deep Convolutional Neural Networks. Computers and Electronics in Agriculture 163, 104859.
  22. Ronneberger, O., F. Philipp, and B. Thomas, 2015: U-Net: Convolutional Networks for Biomedical Image Segmentation. Computer Vision and Pattern Recognition, arXiv:1505.04597
  23. Sang, W. G., J. H. Kim, J. K. Baek, D.W. Kwon, H. Y. Ban, J. I. Cho, and M. C. Seo, 2021: Detection of Drought Stress in Soybean Plants using RGB-based Vegetation Indices. Korean Journal of Agricultural and Forest Meteorology 23(4), 340-348. (in Korean with English abstract)
  24. Lin, T. Y., P. Dollar, R. Girshick, K. He, B. Hariharan, and S. Belongie, 2017: Feature Pyramid Networks for Object Detection. Computer Vision and Pattern Recognition, arXiv:1612.03144
  25. USDA (United States Department of Agriculture), 2020: World Agricultural Production.
  26. Wang, A., W. Zhang, and X. Wei, 2019: A Review on Weed Detection Using Ground-based Machine Vision and Image Processing Techniques. Computers and Electronics in Agriculture 158, 226-240. https://doi.org/10.1016/j.compag.2019.02.005
  27. Xu, W., W. Yang, S. Chen, C. Wu, P. Chen, and Y. Lan, 2020: Establishing a Model to Predict the Single Boll Weight of Cotton in Northern Xinjiang by Using High Resolution UAV Remote Sensing Data. Computers and Electronics in Agriculture 179, 105762.
  28. Zhang, C., K. Zou, and Y. Pan, 2020: A Method of Apple Image Segmentation Based on Color-Texture Fusion Feature and Machine Learning. Agronomy 10(7), 972.
  29. Zhou, D., M. Li, Y. Li, J. Qi, K. Liu, X. Cong, and X. Tian, 2020: Detection of Ground Straw Coverage under Conservation Tillage Based on Deep Learning. Computers and Electronics in Agriculture 172, 105369.
  30. Zou, K., X. Chen, F. Zhang, H. Zhou, and C. Zhang, 2021: A Field Weed Density Evaluation Method Based on UAV Imaging and Modified U-Net. Remote Sensing 13, 310.