Extraction of Film-Mulched Tobacco Fields and Estimation of Tobacco Planting Area Based on Deep Learning and High-Resolution Remote Sensing Images
Information Technology and Control · 03.04.2026 · 55 (1) · 65-85
DOI: 10.5755/j01.itc.55.1.42839 ↗Kurz gefasst
Fachartikel in der Zeitschrift Information Technology and Control, online erschienen am 03.04.2026. Verfasst von Fenghua Huang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref).
Abstract (Originalsprache)
In tobacco cultivation, the use of plastic film mulch in tobacco fields serves as a reliable indicator of farmers’ intent to transplant seedlings from greenhouses to the fields. An important research challenge is how to efficiently and accurately identify film-mulched tobacco fields over large areas using high-resolution satellite remote sensing imagery prior to the transplanting stage. Such identification can support the estimation of actual tobacco planting area, thereby assisting tobacco management authorities in evaluating the fulfillment of macro-level planting targets and formulating regulatory policies. To address the limitations of conventional manual and machine learning methods, such as low efficiency and insufficient accuracy in extracting tobacco field boundaries from high-resolution remote sensing images, this study proposes an approach based on the Skip_Segformer semantic segmentation model. Specifically, a SKIPAT module was integrated into the encoder of the traditional SegFormer model to reduce the number of training parameters and save computational resources. Additionally, the decoder was enhanced with a Multi-level Feature Fusion (MFF) mechanism to better integrate features across different scales, thereby significantly improving the accuracy of film-mulched field boundary extraction. The experiment was conducted in Siqian Township, Guangze County, Nanping City, using Jilin-1 satellite imagery with a spatial resolution of 0.5 meters, acquired on March 8, 2022. At both full-regional and local scales, the Skip_Segformer model was compared with four other networks (DeeplabV3+, Hrnet, Pspnet, and SegFormer) in extracting film-mulched tobacco field patches. The results were compared to identify the optimal model. Experimental results demonstrate that the Skip_Segformer achieved the highest extraction accuracy and generalization capability among the compared models. It attained an extraction accuracy of 97% across Siqian Township, with a relative error of only 1.6% in the estimated tobacco planting area, significantly outperforming the other four models. The proposed method shows strong feasibility and applicability for large-scale extraction of film-covered tobacco fields and estimation of planting area, effectively supporting tobacco administration departments in total planting area monitoring and providing a basis for local tobacco planting planning.
Zitieren
APA
Huang, F., Gao, R., Wang, L., Zhao, Q., & Chi, G. (2026). Extraction of Film-Mulched Tobacco Fields and Estimation of Tobacco Planting Area Based on Deep Learning and High-Resolution Remote Sensing Images. Information Technology and Control, 55(1), 65-85. https://doi.org/10.5755/j01.itc.55.1.42839
Vancouver
Huang F, Gao R, Wang L, Zhao Q, Chi G. Extraction of Film-Mulched Tobacco Fields and Estimation of Tobacco Planting Area Based on Deep Learning and High-Resolution Remote Sensing Images. Information Technology and Control. 2026;55(1):65-85. doi:10.5755/j01.itc.55.1.42839
BibTeX
@article{Huang2026,
author = {Huang, Fenghua and Gao, Ronggang and Wang, Lin and Zhao, Qianyu and Chi, Guosheng},
title = {Extraction of Film-Mulched Tobacco Fields and Estimation of Tobacco Planting Area Based on Deep Learning and High-Resolution Remote Sensing Images},
journal = {Information Technology and Control},
year = {2026},
volume = {55},
number = {1},
pages = {65--85},
doi = {10.5755/j01.itc.55.1.42839}
}
EUVIMED