All Issue

2024 Vol.13, Issue 4 Preview Page

Research Article

Detection of Abandoned Fields in Aerial Images using Deep Learning (YOLOv5)

YOLOv5를 이용한 항공 영상에서 폐경지 후보지 탐지

최호연1, 홍지화2, 김용현3, 송지숙4, 박은수5†

Hoyeon Choi1, Jeehwa Hong2, Yonghyun Kim3, Jisook Song4, Eunsoo Park5†

1국립농산물품질관리원 시험연구소 품질조사과 보조원
2국립농산물품질관리원 시험연구소 품질조사과 연구사
3국립농산물품질관리원 시험연구소 품질조사과 과장
4국립농산물품질관리원 시험연구소 소장
5국립농산물품질관리원 시험연구소 품질조사과 연구사
1Research Assistant, Quality Examination Division, Experiment Research Institute, National Agricultural Products Quality Management Service
2Researcher, Quality Examination Division, Experiment Research Institute, National Agricultural Products Quality Management Service
3Director of Quality Examination Division, Quality Examination Division, Experiment Research Institute, National Agricultural Products Quality Management Service
4Director of General of Experiment Research Institute, Experiment Research Institute, National Agricultural Products Quality Management Service
5Researcher, Quality Examination Division, Experiment Research Institute, National Agricultural Products Quality Management Service
31 December 2024. pp. 499~507
PDF Citation
Abstract
The Ministry of Agriculture, Food and Rural Affairs operates Public Direct Payment Program that provides subsidies to farmers who create public interest functions through agriculture activities, thereby encouraging the production, maintenance, and expansion of public interest values. The domestic Agriculture Business Entities management work verifies whether there are abandoned fields (If not operated as farmland, direct payments will be reduced.) based on aerial images, and then verifies whether there is abandoned field through on-site inspection. This study aimed to detect abandoned fields in various agricultural lands by utilizing ‘The you only look once version 5(YOLOv5)’ model, which is effective in object detection of images based on aerial images. We collected aerial images of various areas targeting four objects that frequently appear in agricultural lands: graves, solar panels, buildings, and greenhouses. After cropping the aerial images into 1024x1024 resolution sizes from their actual sizes, labeling (box annotation) was performed. The analysis results showed an acceptable average recall (AP) of approximately 0.739, and further precision improvement is expected if additional data is collected. If abandoned fields are detected through connection with the land register map in the future, it is expected that direct payment work can be performed efficiently.
Keywords
Public direct payment
Abandoned fields
Aerial images
Object detection
YOLOv5
농림축산식품부는 농업의 공익기능 증진과 농가 소득 안정을 위해 공익직불제를 운영하며, 보조금(공익지불금) 지급 대상은 농업경영체에 등록된 농지로, ‘농업농촌공익직불법’에 따라 농지의 형상과 기능이 유지되어야 한다. 직불금 관리는 항공영상으로 폐경지(농지로 이용이 불가능하게 농지 형상이 변한 농경지)를 검토한 뒤 현장 확인을 통해 검증하며, 이는 많은 시간과 노동력이 소요된다. 본 연구는 항공영상을 기반으로 영상의 객체인식(Object Detection) 분야에서 효과적인 성능을 발휘하는 You Only Look Once Version 5(YOLOv5) 모델을 활용하여 다양한 농지의 폐경지 후보지를 탐지하고자 하였다. 대상을 농지에서 폐경지로 많이 나타나는 건물, 묘지, 온실, 태양광 시설과 같은 4가지를 대상으로 다양한 지역의 항공영상을 수집하였다. 항공영상의 실제 크기에서 1024x1024 해상도 크기로 분할한 후 라벨링(box annotation)을 수행하였다. 분석 결과, 평균 재현율(Average Recall, AR)이 약 0.739로 준수한 결과가 나타났다. 개발된 모델을 활용한다면, 직불 업무 중 항공 영상에서 폐경지 후보지 확인 단계에서 효율적으로 활용할 수 있을 것이라 판단된다. 또한 지속적으로 폐경지 후보 유형을 추가하고, GIS와 연계한다면, 정확도 향상과 자동 판단을 통한 업무 자동화가 이루어질 것이라 예상된다.
키워드
공익직불금
폐경지
항공영상
객체인식
YOLOv5
References
  1. 김하영･나라･주동혁･최규훈･오윤경, 2022, “YOLO v2를 이용한 고해상도 항공영상에서의 태양광발전소 탐지 방법 연구,” 농촌계획, 28(2), 87-96.10.7851/ksrp.2022.28.2.087
  2. 박강현･윤유정･강종구･김근아･최소연･장선웅･박수호･공신우･곽지우･이양원, 2022, “YOLOv5와 YOLOv7 모델을 이용한 해양침적쓰레기 객체탐지 비교평가,” 대한원격탐사학회지, 38(6), 1643-1652.
  3. 박기홍･홍일영, 2017, “보급형 회전익 무인항공기를 이용한 토지행정 지원 연구,” 한국지리학회지, 6(1), 93-100.10.25202/JAKG.6.1.7
  4. 임윤교･윤유정･강종구･김서연･정예민･최소연･서영민･이양원, 2023, “YOLO를 이용한 SAR 영상의 선박 객체 탐지: 편파별 모델 구성과 정확도 특성 분석,” 대한원격탐사학회지, 39(5), 997-1008.
  5. Alrasheedi, F., Zhong, X., and Huang, P.C., 2023, Padding module: Learning the padding in deep neural networks, IEEE Access, 11.10.1109/ACCESS.2023.3238315
  6. Girshick, R., Donahue, J., Darrell, T., and Malik, J., 2014, Rich feature hierarchies for accurate object detection and semantic segmentation, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, June 24-27, Ohio, United States, 580-587.10.1109/CVPR.2014.81
  7. Huang, G., Liu, Z., Van Der Maaten, L., and Weinberger, K.Q., 2017, Densely connected convolutional networks, Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, July 22-25, Hawail, USA, 4700-4708.10.1109/CVPR.2017.243PMC5598342
  8. Kim, J.S. and Hong, Y., 2021, Analysis of building object detection based on the YOLO neural network using UAV images, Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography, 39(6), 361-392.
  9. Parhar, P., Sawasaki, R., Todeschini, A., Vahabi, H., Nusaputra, N., and Vergara, F., 2022, HyperionSolarNet: Solar panel detection from aerial images, ArXiv Preprint ArXiv:2201.02107.
  10. Park, C.W. and Jung, H.S., 2022, Detection of urban trees using YOLOv5 from aerial images, Korean Journal of Remote Sensing, 38(6), 1633-1641.
  11. Park, G., Youn, Y., Kang, J., Kim, G., Choi, S., Jang, S., Bak, S., Gong, S., Kwak, J., and Lee, Y. 2022, A comparative study on the object detection of deposited marine debris (DMD) using YOLOv5 and YOLOv7 models, Korean Journal of Remote Sensing, 38(6), 1643-1652.
  12. Pavlov, V.A. and Galeeva, M.A., 2019, Detection and recognition of objects on aerial photographs using convolutional neural networks, Journal of Physics: Conference Series, 1326(1), 12038.10.1088/1742-6596/1326/1/012038
  13. Redmon, J., Divvala, S., Girshick, R., and Farhadi, A., 2016, You only look once: Unified, real-time object detection, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, June 27-30, Lasvegas, USA, 779-788.10.1109/CVPR.2016.91
  14. Santos, C.F.G.D. and Papa, J.P., 2022, Avoiding overfitting: A survey on regularization methods for convolutional neural networks, In ACM Computing Surveys, 54(10), 1-25.10.1145/3510413
  15. Shreyas Dixit, K.G., Chadaga, M.G., Savalgimath, S.S., Ragavendra Rakshith, G., and Naveen Kumar, M.R., 2019, Evaluation and evolution of object detection techniques YOLO and R-CNN, International Journal of Recent Technology and Engineering, 8(2S3).10.35940/ijrte.B1154.0782S319
  16. Terven, J. and Cordova-Esparza, D., 2023, A comprehensive review of YOLO: From YOLOv8 and beyond, Machine learning & knowledge extraction, 5, 1680-1716.10.3390/make5040083
  17. Xu, Y., Wu, L., Xie, Z., and Chen, Z., 2018, Building extraction in very high resolution remote sensing imagery using deep learning and guided filters, Remote Sensing, 10(1), 144.10.3390/rs10010144
  18. Yoo, S., Kim, C. H., Kwon, Y., Choi, W., and Sohn, H.G., 2022, Automatic building extraction using SpaceNet building dataset and context-based ResU-Net, Korean Journal of Remote Sensing, 38(5-2), 685-694.
  19. Yoon, B., Seong, S., and Choi, J, 2023, Detection of plastic greenhouses by using deep learning model for aerial orthoimages, Korean Journal of Remote Sensing, 39(2), 183-192.
  20. 국가법령정보센터, 농업･농촌 공익기능 증진 직접지불제도 운영에 관한 법률, https://www.law.go.kr
  21. 통계청, 2023년 농경지 면적 조사 결과, https://kostat.go.kr/anse/
  22. 한국농촌경제연구원, 휴경농지 현황과 정책 방향, https://repository.krei.re.kr/ handle/2018.oak/14918
  23. CS231n, Solar Panel Detection on Satellite Images: From Faster R-CNN to YOLOv10, https://cs231n.stanford.edu/2024/reports.html
  24. Github, YOLOv5, https://github.com/ultralytics/yolov5
Information
  • Publisher :The Association of Korean Geographers
  • Publisher(Ko) :한국지리학회
  • Journal Title :Journal of the Association of Korean Geographers
  • Journal Title(Ko) :한국지리학회지
  • Volume : 13
  • No :4
  • Pages :499~507
  • DOI :https://doi.org/10.25202/JAKG.13.4.11
Journal Informaiton Journal of the Association of Korean Geographers Journal of the Association of Korean Geographers
Online Submission
  • NRF
  • KISTI Current Status
  • KISTI Cited-by
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close