농업지 모니터링을 위한 2단계 U-Net 기반이진분할 정확도 향상 방법 (Improving Binary Segmentation Accuracy for Agricultural Land Monitoring through Two-Stage U-Net Training)

안향식 (HyangSig Ahn); 신영규 (YoungKyu Shin); 최태봉 (TaeBong Choi)

doi:10.14249/eia.2025.34.6.624

Improving Binary Segmentation Accuracy for Agricultural Land Monitoring through Two-Stage U-Net Training

Journal of Environmental Impact Assessment
Abbr : J EIA
2025, 34(6), pp.624~641
DOI : 10.14249/eia.2025.34.6.624
Publisher : Korean Society Of Environmental Impact Assessment
Research Area : Engineering > Environmental Engineering
Received : December 4, 2025
Accepted : December 18, 2025
Published : December 30, 2025

HyangSig Ahn ¹, YoungKyu Shin ¹, TaeBong Choi ¹

¹국립환경과학원

Accredited

ABSTRACT

Binary segmentation faces persistent challenges from false negatives (FNs), particularly inapplications where missing small target regions can undermine reliability. FNs are critical in sparseland-cover mapping tasks because they represent true targets that models fail to detect. Here, wepropose a two-stage U-Net training strategy that automatically accumulates FP and FN regions fromStage 1 predictions and reuses them as auxiliary supervisory signals through a second output headduring Stage 2 learning. This enhances the model's ability to revisit uncertain or ambiguous areas without any architectural modification. The method was applied to mapping sparsely distributedginseng fields from high-resolution aerial imagery from South Korea. The approach substantiallyreduced FNs, tolerated a modest increase in FPs, and improved Intersection over Union (IoU) byapproximately 0.10 across various training conditions. Because the strategy only modifies the trainingscheme and not the model architecture, it is compatible with U-Net variants and generalizable to otherdomains where FN suppression and stability are essential. The findings demonstrate that learningfrom accumulated error regions offers an efficient and practical pathway toward more reliablesegmentation in complex real world environments.

KEYWORDS

aerial imagery, two-stage training, false negatives, ginseng field mapping, agricultural land cover segmentation

Journal Copyright Policy

No CCL information provided

Citation status

* References for papers published after 2025 are currently being built.

[confproc] Abraham, S. / 2019 / A Novel Focal Tversky Loss Function with Improved Attention U-Net for Lesion Segmentation / Proceedings of the IEEE International Symposium on Biomedical Imaging (ISBI 2019)

[journal] Cardama, F. J. / 2023 / Consensus Techniques for Unsupervised Binary Change Detection Using Multi-Scale Segmentation Detectors for Land Cover Vegetation Images / Remote Sensing 15 : 2889~

[journal] Caruana, R. / 1997 / Multitask Learning / Machine Learning 28 : 41~75

[journal] Chen, L. C. / 2018 / DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs / IEEE Transactions on Pattern Analysis and Machine Intelligence 40 : 834~848

[journal] Cheng H. / 2025 / MSGU-Net: A Lightweight Multi-Scale Ghost U-Net for Image Segmentation / Frontier Neurorobot 18 : 1480055~

[journal] Chen H. / 2024 / MS-UNet: Multi-Scale Nested UNet for Medical Image Segmentation with Few Training Data Based on an ELoss and Adaptive Denoising Method / Mathematics 12 : 2996~

[journal] Chen, Z. / 2022 / A Convolutional Neural Network for Large-Scale Greenhouse Extraction from Satellite Images Considering Spatial Features / Remote Sensing 14 : 4908~

[confproc] Fan D. P. / 2020 / PraNet: Parallel Reverse Attention Network for Polyp Segmentation / Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2020)

[book] Goodfellow, I. / 2016 / Deep Learning / MIT Press

[journal] Huang, Y. / 2020 / Batching Soft IoU for Training Semantic Segmentation Networks / IEEE Signal Processing Letters 27 : 66~70

[confproc] Jha, D. / 2020 / Double U-Net: A Deep Convolutional Neural Network for Medical Image Segmentation / Proceedings of the 33rd IEEE International Symposium on Computer-Based Medical Systems (CBMS 2020)

[confproc] Jha D. / 2019 / ResUNet++: An Advanced Architecture for Medical Image Segmentation / Proceedings of the 2019 IEEE International Symposium on Multimedia (ISM 2019)

[confproc] Jiang, Z. / 2019 / Two-Stage Cascaded U-Net: 1st Place Solution to BraTS Challenge 2019 Segmentation Task / Proceedings of the International MICCAI BrainLesion Workshop (BrainLes 2019)

[other] Kingma, D. P. / 2014 / Adam: A Method for Stochastic Optimization / arXiv

[journal] Lee, D. H. / 2021 / UAV, a Farm Map, and Machine Learning Technology Convergence Classification Method of a Corn Cultivation Area / Agronomy 11 : 1554~

[journal] Lee, K. / 2024 / Fine-Grained Binary Segmentation for Geospatial Objects in Remote Sensing Imagery via Path-Selective Test-Time Adaptation / IEEE Trans. Geosci. Remote Sensing 62 : 5623216~

[confproc] Lin, T. Y. / 2017 / Focal Loss for Dense Object Detection / Proceedings of the IEEE International Conference on Computer Vision (ICCV 2017)

[confproc] Maggiori, E. / 2017 / Can semantic labeling methods generalize to any city? The INRIA aerial image labeling benchmark / Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS)

[other] Oktay, O. / 2018 / Attention U-Net: Learning Where to Look for the Pancreas / arXiv

[journal] Pang, B. / 2024 / GAUNet: A Lightweight Ghost and Attention U-Net for Medical Image Segmentation / Journal of Imaging Informatics in Medicine 37 : 1874~1888

[confproc] Paszke, A. / 2019 / PyTorch: An Imperative Style, High-Performance Deep Learning Library / Advances in Neural Information Processing Systems (NeurIPS) / Curran Associates, Inc : 8024~8035

[web] Python Software Foundation / 2024 / Python: A Programming Language / Python Software Foundation / https://www.python.org/

[journal] Qin, X. / 2020 / U²-Net: Going Deeper with Nested U-Structure for Salient Object Detection / Pattern Recognition 106 : 107404~

[confproc] Qin X. / 2019 / BASNet: Boundary-Aware Salient Object Detection / Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2019)

[confproc] Rahman, M. A. / 2016 / Optimizing Intersectionover-Union in Deep Neural Networks for Image Segmentation / Proceedings of the International Symposium on Visual Computing (ISVC 2016)

[confproc] Ronneberger, O. / 2015 / U-Net: Convolutional Networks for Biomedical Image Segmentation / Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2015)

[confproc] Rottmann, M. / 2023 / Automated Detection of Label Errors in Semantic Segmentation Datasets via Deep Learning and Uncertainty Quantification / Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

[journal] Rumelhart, D. E. / 1986 / Learning Representations by Back-Propagating Errors / Nature 323 : 533~536

[other] Salman, M. S. / 2019 / Overfitting Mechanism and Avoidance in Deep Neural Networks / arXiv

[confproc] Schubert, M. / 2024 / Identifying label errors in object detection datasets by loss inspection / Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

[journal] Shorten, C. / 2019 / Survey on Image Data Augmentation for Deep Learning / Jounal of Big Data 6 : 1~48

[journal] Srivastava, N. / 2014 / Dropout: A Simple Way to Prevent Neural Networks from Overfitting / Journal of Machine Learning. Research 15 : 1929~1958

[journal] Yang Z. / 2021 / Mapping Panax notoginseng Plantations by Using an Integrated Pixel-and Object-Based (IPOB) Approach and ZY-3 Imagery / Remote Sensing 13 : 2184~

[confproc] Zhou, Z. / 2018 / UNet++: A Nested U-Net Architecture for Medical Image Segmentation / Proceedings of the International Workshop on Deep Learning in Medical Image Analysis

KJCKorea
Journal Central

Journal of Environmental Impact Assessment 2025 KCI Impact Factor : 0.63

Improving Binary Segmentation Accuracy for Agricultural Land Monitoring through Two-Stage U-Net Training

ABSTRACT

KEYWORDS

Journal Copyright Policy

Citation status

* References for papers published after 2025 are currently being built.

Journal of Environmental Impact Assessment 2025 KCI Impact Factor : 0.63

Improving Binary Segmentation Accuracy for Agricultural Land Monitoring through Two-Stage U-Net Training

ABSTRACT

KEYWORDS

Journal Copyright Policy

Statistics

Tools

Issue List

Citation status

KCI Citation Counts (0)

REFERENCES (34) * References for papers published after 2025 are currently being built.

Search PDF

Citation

* References for papers published after 2025 are currently being built.