Enhanced Garlic Crop Identification Using Deep Learning Edge Detection and Multi-Source Feature Optimization with Random Forest
Abstract
1. Introduction
- Deep Learning-Based Boundary Extraction: Implementation of a DexiNed network on high-resolution imagery to accurately extract and vectorize agricultural field boundaries, addressing spatial fragmentation issues in traditional pixel-based crop classification approaches.
- Optimized Multi-Source Feature Fusion: Integration of Sentinel-1 SAR and Sentinel-2 optical time-series data to construct comprehensive feature sets encompassing polarimetric, spectral, vegetation index, and textural characteristics, with systematic feature selection combining Random Forest Importance and Recursive Feature Elimination techniques.
- Field-Level Spatial Constraints: Incorporation of extracted agricultural boundaries as spatial constraints to refine pixel-level classification results and generate field-aligned garlic distribution maps that align with actual agricultural production units.
- Comprehensive Framework Validation: Demonstration of the integrated approach’s effectiveness in achieving high mapping accuracy, reliable multi-source data integration, and practical applicability for precision agriculture applications in complex agricultural landscapes.
2. Materials and Methods
2.1. Study Area
2.2. Sentinel-1/2 Data
2.3. Cultivated Land Field Data
2.4. Sample Data
2.5. Feature Variables
2.6. DexiNed
2.7. Feature Optimization
2.8. Random Forest
2.9. Evaluation Criteria
2.10. Workflow of the Study
3. Results
3.1. Optimized Hyperparameter Results for Random Forest
3.2. Multi-Feature Optimization Results
3.3. Random Forest Classification Results
3.4. Agricultural Land Boundary Extraction Based on Deep Learning
3.5. Optimization of Classification Results Based on Field Constraints
4. Discussion
4.1. Methodological Innovation Through Technical Integration
4.2. Integrated Technical Performance Analysis
4.3. Methodological Limitations and Future Improvements
5. Conclusions
- An efficient feature optimization strategy was established: Recursive Feature Elimination (RFE) selected 13 optimal features from 28 candidates, improving classification accuracy from 91% to 93%, validating the critical role of feature selection in remote sensing classification; vegetation indices contributed critically to garlic identification, with GNDVI, NDVI, and OSAVI exhibiting significantly higher importance than original spectral bands;
- Deep learning was successfully applied for agricultural field extraction: the DexiNed edge detection network demonstrated outstanding performance in agricultural field boundary delineation, achieving User’s Accuracy, Producer’s Accuracy, and F1-score of 93.22%, 95.90%, and 94.16%, respectively, providing a reliable spatial constraint foundation for subsequent field-based crop identification;
- An effective spatially constrained optimization strategy was proposed: the majority voting optimization method based on field boundaries successfully resolved salt-and-pepper noise in pixel-level classification, significantly enhancing the spatial continuity and integrity of classification results, rendering the final products better aligned with actual agricultural production units;
- High-precision garlic identification was achieved: application in major garlic-producing areas of Kaifeng City demonstrated the method’s capability to accurately identify garlic cultivation zones, providing technical support for monitoring and management within the garlic industry.
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Zhong, Y.; Lu, F.; Li, C.; Qiao, Q.; Tang, Z.; Duan, Z.; Yu, Y.; Wang, T.; Xie, D. Analysis of the Development and Trade Dynamics of the Global Garlic Industry. China Veg. 2025, 1, 1–17. [Google Scholar] [CrossRef]
- Komaraasih, R.I.; Sitanggang, I.S.; Agmalaro, M.A. Sentinel-1A Image Classification for Identification of Garlic Plants Using a Decision Tree Algorithm. In Proceedings of the 2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA), Bogor, Indonesia, 16–17 September 2020; pp. 1–6. [Google Scholar]
- Tian, H.; Pei, J.; Huang, J.; Li, X.; Wang, J.; Zhou, B.; Qin, Y.; Wang, L. Garlic and Winter Wheat Identification Based on Active and Passive Satellite Imagery and the Google Earth Engine in Northern China. Remote Sens. 2020, 12, 3539. [Google Scholar] [CrossRef]
- Rahmati, A.; Zoej, M.J.V.; Dehkordi, A.T. Early Identification of Crop Types Using Sentinel-2 Satellite Images and an Incremental Multi-Feature Ensemble Method (Case Study: Shahriar, Iran). Adv. Space Res. 2022, 70, 907–922. [Google Scholar] [CrossRef]
- Yang, S.; Gu, L.; Li, X.; Jiang, T.; Ren, R. Crop Classification Method Based on Optimal Feature Selection and Hybrid CNN-RF Networks for Multi-Temporal Remote Sensing Imagery. Remote Sens. 2020, 12, 3119. [Google Scholar] [CrossRef]
- Chauhan, H.; Krishna Mohan, B. Effectiveness of Spectral Similarity Measures to Develop Precise Crop Spectra for Hyperspectral Data Analysis. ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci. 2014, II-8, 83–90. [Google Scholar] [CrossRef]
- Wu, Y.; Wu, Y.; Wang, B.; Yang, H. A Remote Sensing Method for Crop Mapping Based on Multiscale Neighborhood Feature Extraction. Remote Sens. 2022, 15, 47. [Google Scholar] [CrossRef]
- Belgiu, M.; Csillik, O. Sentinel-2 Cropland Mapping Using Pixel-Based and Object-Based Time-Weighted Dynamic Time Warping Analysis. Remote Sens. Environ. 2018, 204, 509–523. [Google Scholar] [CrossRef]
- Li, G.; Cui, J.; Han, W.; Zhang, H.; Huang, S.; Chen, H.; Ao, J. Crop Type Mapping Using Time-Series Sentinel-2 Imagery and U-Net in Early Growth Periods in the Hetao Irrigation District in China. Comput. Electron. Agric. 2022, 203, 107478. [Google Scholar] [CrossRef]
- Chen, Y.; Guo, Y.; Qiao, L.; Xia, H. Coupling Optical and SAR Imagery for Automatic Garlic Mapping. Front. Sustain. Food Syst. 2022, 6, 1007568. [Google Scholar] [CrossRef]
- Zhong, L.; Hu, L.; Zhou, H. Deep Learning Based Multi-Temporal Crop Classification. Remote Sens. Environ. 2019, 221, 430–443. [Google Scholar] [CrossRef]
- Van Tricht, K.; Gobin, A.; Gilliams, S.; Piccard, I. Synergistic Use of Radar Sentinel-1 and Optical Sentinel-2 Imagery for Crop Mapping: A Case Study for Belgium. Remote Sens. 2018, 10, 1642. [Google Scholar] [CrossRef]
- Niu, Y.; Zhang, L.; Zhang, H.; Han, W.; Peng, X. Estimating Above-Ground Biomass of Maize Using Features Derived from UAV-Based RGB Imagery. Remote Sens. 2019, 11, 1261. [Google Scholar] [CrossRef]
- Lu, T.; Gao, M.; Wang, L. Crop Classification in High-Resolution Remote Sensing Images Based on Multi-Scale Feature Fusion Semantic Segmentation Model. Front. Plant Sci. 2023, 14, 1196634. [Google Scholar] [CrossRef]
- Wang, H.; Chang, W.; Yao, Y.; Yao, Z.; Zhao, Y.; Li, S.; Liu, Z.; Zhang, X. Cropformer: A New Generalized Deep Learning Classification Approach for Multi-Scenario Crop Classification. Front. Plant Sci. 2023, 14, 1130659. [Google Scholar] [CrossRef]
- Belgiu, M.; Drăguţ, L. Random Forest in Remote Sensing: A Review of Applications and Future Directions. ISPRS J. Photogramm. Remote Sens. 2016, 114, 24–31. [Google Scholar] [CrossRef]
- You, N.; Dong, J.; Huang, J.; Du, G.; Zhang, G.; He, Y.; Yang, T.; Di, Y.; Xiao, X. The 10-m Crop Type Maps in Northeast China during 2017–2019. Sci. Data 2021, 8, 41. [Google Scholar] [CrossRef] [PubMed]
- Ma, Z.; Xue, H.; Liu, C.; Li, C.; Fang, X.; Zhou, J. Identification of Garlic Based on Active and Passive Remote Sensing Data and Object-Oriented Technology. Trans. Chin. Soc. Agric. Eng. 2022, 38, 210–222. [Google Scholar]
- Xu, L.; Yang, P.; Yu, J.; Peng, F.; Xu, J.; Song, S.; Wu, Y. Extraction of Cropland Field Parcels with High Resolution Remote Sensing Using Multi-Task Learning. Eur. J. Remote Sens. 2023, 56, 2181874. [Google Scholar] [CrossRef]
- Genuer, R.; Poggi, J.-M.; Tuleau-Malot, C. Variable Selection Using Random Forests. Pattern Recognit. Lett. 2010, 31, 2225–2236. [Google Scholar] [CrossRef]
- Zhu, Y.; Pan, Y.; Hu, T.; Zhang, D.; Zhao, C.; Gao, Y. A Generalized Framework for Agricultural Field Delineation from High-Resolution Satellite Imageries. Int. J. Digit. Earth 2024, 17, 2297947. [Google Scholar] [CrossRef]
- Soria, X.; Sappa, A.; Humanante, P.; Akbarinia, A. Dense Extreme Inception Network for Edge Detection. Pattern Recognit. 2023, 139, 109461. [Google Scholar] [CrossRef]
- Kursa, M.B.; Rudnicki, W.R. Feature Selection with the Boruta Package. J. Stat. Softw. 2010, 36, 1–13. [Google Scholar] [CrossRef]
- Kussul, N.; Lavreniuk, M.; Skakun, S.; Shelestov, A. Deep Learning Classification of Land Cover and Crop Types Using Remote Sensing Data. IEEE Geosci. Remote Sens. Lett. 2017, 14, 778–782. [Google Scholar] [CrossRef]
- Rodriguez-Galiano, V.F.; Ghimire, B.; Rogan, J.; Chica-Olmo, M.; Rigol-Sanchez, J.P. An Assessment of the Effectiveness of a Random Forest Classifier for Land-Cover Classification. ISPRS J. Photogramm. Remote Sens. 2012, 67, 93–104. [Google Scholar] [CrossRef]
- Waldner, F.; Canto, G.S.; Defourny, P. Automated Annual Cropland Mapping Using Knowledge-Based Temporal Features. ISPRS J. Photogramm. Remote Sens. 2015, 110, 1–13. [Google Scholar] [CrossRef]
- Gao, Y.; Zhao, Z.; Shang, G.; Liu, Y.; Liu, S.; Yan, H.; Chen, Y.; Zhang, X.; Li, W. Optimal Feature Selection and Crop Extraction Using Random Forest Based on GF-6 WFV Data. Int. J. Remote Sens. 2024, 45, 7395–7414. [Google Scholar] [CrossRef]
- Yin, L.; You, N.; Zhang, G.; Huang, J.; Dong, J. Optimizing Feature Selection of Individual Crop Types for Improved Crop Mapping. Remote Sens. 2020, 12, 162. [Google Scholar] [CrossRef]
- Trivedi, M.B.; Marshall, M.; Estes, L.; De Bie, C.A.J.M.; Chang, L.; Nelson, A. Cropland Mapping in Tropical Smallholder Systems with Seasonally Stratified Sentinel-1 and Sentinel-2 Spectral and Textural Features. Remote Sens. 2023, 15, 3014. [Google Scholar] [CrossRef]
- Kumar, S.S.; Shaikh, T. Empirical Evaluation of the Performance of Feature Selection Approaches on Random Forest. In Proceedings of the 2017 International Conference on Computer and Applications (ICCA), Doha, Qatar, 6–7 September 2017; pp. 227–231. [Google Scholar]
- Tariq, A.; Yan, J.; Gagnon, A.S.; Riaz Khan, M.; Mumtaz, F. Mapping of Cropland, Cropping Patterns and Crop Types by Combining Optical Remote Sensing Images with Decision Tree Classifier and Random Forest. Geo-Spat. Inf. Sci. 2023, 26, 302–320. [Google Scholar] [CrossRef]
- Waldner, F.; Diakogiannis, F.I. Deep Learning on Edge: Extracting Field Boundaries from Satellite Images with a Convolutional Neural Network. Remote Sens. Environ. 2020, 245, 111741. [Google Scholar] [CrossRef]
- Kordi, F.; Yousefi, H. Crop Classification Based on Phenology Information by Using Time Series of Optical and Synthetic-Aperture Radar Images. Remote Sens. Appl. Soc. Environ. 2022, 27, 100812. [Google Scholar] [CrossRef]
Center Wavelength (nm) | ||||
---|---|---|---|---|
Transducer | Band Name | S2A | S2B | Resolution (m) |
Sentinel-1 SAR | VV | 5.405 GHz 5.405 GHz | 10 | |
VH | 10 | |||
B1 | 443.9 | 442.3 | 60 | |
Sentinel-2 MSI | B2 | 496.6 | 492.1 | 10 |
B3 | 560.0 | 559.0 | 10 | |
B4 | 664.5 | 665.0 | 10 | |
B5 | 703.9 | 703.8 | 20 | |
B6 | 740.2 | 739.1 | 20 | |
B7 | 782.5 | 779.7 | 20 | |
B8 | 835.1 | 833.0 | 10 | |
B8A | 864.8 | 864.0 | 20 | |
B9 | 945.0 | 943.2 | 60 | |
B10 | 1373.5 | 1376.9 | 60 | |
B11 | 1613.7 | 1610.4 | 20 | |
B12 | 2202.4 | 2185.7 | 20 |
Data Category | Winter Wheat | Garlic | Other Land | Building | Water | Total |
---|---|---|---|---|---|---|
Training datasets | 747 | 417 | 74 | 245 | 93 | 1576 |
Verification datasets | 321 | 180 | 30 | 106 | 39 | 676 |
Total | 1068 | 597 | 104 | 351 | 132 | 2252 |
Sensor | Feature Type | Feature Variables |
---|---|---|
Sentinel-1 | Polarization Features | VV |
VH | ||
Sentinel-2 | Spectral Features | B2 |
B3 | ||
B4 | ||
B5 | ||
B6 | ||
B8 | ||
B8A | ||
B11 | ||
B12 | ||
Vegetation index Features | Normalized Difference Vegetation Index (NDVI) | |
Enhanced Vegetation Index (EVI) | ||
Soil-Adjusted Vegetation Index (SAVI) | ||
Green Normalized Difference Vegetation Index (GNDVI) | ||
Green Leaf Index (GLI) | ||
Optimized Soil-Corrected Vegetation Index (OSAVI) | ||
Ratio Vegetation Index (RVI) | ||
Normalized Difference Red Edge (NDRE) | ||
Normalized Difference Built-up Index (NDBI) | ||
Normalized Difference Water Index (NDWI) | ||
Normalized Burn Ratio (NBR) | ||
Modified Normalized Difference Water Index (MNDWI) | ||
Texture features | NIR Variance Texture Feature (Nir_variance) | |
NIR Energy Texture Feature (Nir_energy) | ||
NIR Entropy Texture Feature (Nir_entropy) NIR Homogeneity Texture Feature (Nir_homogeneity) |
Ground Truth | |||
---|---|---|---|
Positive | Negative | ||
Prediction | Positive | ||
Negative |
Classification Category | Before Feature Selection | After Feature Selection | ||||
---|---|---|---|---|---|---|
UA | PA | F1-Score | UA | PA | F1-Score | |
Wheat | 0.93 | 0.93 | 0.93 | 0.94 ↑0.01 | 0.94 ↑0.01 | 0.94 ↑0.01 |
Garlic | 0.88 | 0.87 | 0.88 | 0.89 ↑0.01 | 0.90 ↑0.03 | 0.90 ↑0.02 |
Others | 0.92 | 0.79 | 0.85 | 0.96 ↑0.04 | 0.86 ↑0.07 | 0.91 ↑0.06 |
Building | 0.89 | 0.93 | 0.91 | 0.95 ↑0.06 | 0.94 ↑0.01 | 0.94 ↑0.03 |
Water | 0.91 | 0.91 | 0.91 | 0.94 ↑0.03 | 0.91 | 0.93 ↑0.02 |
OA = 0.91 Kappa = 0.8654 | OA = 0.93 ↑0.02 Kappa = 0.8857 ↑0.0203 |
Network Model | Evaluation Indicators | ||
---|---|---|---|
UA | PA | F1-Score | |
DexiNed | 93.22% | 95.90% | 94.16% |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Zhou, J.; Diao, Q.; Liu, X.; Su, H.; Yang, Z.; Ma, Z. Enhanced Garlic Crop Identification Using Deep Learning Edge Detection and Multi-Source Feature Optimization with Random Forest. Sensors 2025, 25, 6014. https://doi.org/10.3390/s25196014
Zhou J, Diao Q, Liu X, Su H, Yang Z, Ma Z. Enhanced Garlic Crop Identification Using Deep Learning Edge Detection and Multi-Source Feature Optimization with Random Forest. Sensors. 2025; 25(19):6014. https://doi.org/10.3390/s25196014
Chicago/Turabian StyleZhou, Junli, Quan Diao, Xue Liu, Hang Su, Zhen Yang, and Zhanlin Ma. 2025. "Enhanced Garlic Crop Identification Using Deep Learning Edge Detection and Multi-Source Feature Optimization with Random Forest" Sensors 25, no. 19: 6014. https://doi.org/10.3390/s25196014
APA StyleZhou, J., Diao, Q., Liu, X., Su, H., Yang, Z., & Ma, Z. (2025). Enhanced Garlic Crop Identification Using Deep Learning Edge Detection and Multi-Source Feature Optimization with Random Forest. Sensors, 25(19), 6014. https://doi.org/10.3390/s25196014