基于“分区-分层”深度学习的干旱区土地利用遥感精细制图

doi:10.12118/j.issn.1000-6060.2025.732

Abstract

Abstract:

The pronounced spatial heterogeneity of surface landscapes and the fragmented, intricately mixed distribution of land cover types in arid regions highlight the importance of fine-grained land-use remote-sensing mapping for effective territorial spatial planning and ecological governance. This study focuses on the central urban area of Hami City, Xinjiang, and proposes an improved framework for fine-grained land-use remote-sensing mapping. The framework integrates the geoscientific “zoning-stratification” strategy with advanced deep learning algorithms, utilizing 1-m high-resolution satellite imagery from the ESRI platform. The proposed framework consists of three innovative components: Geographical zoning, stratified object extraction, and post-processing optimization. The results demonstrate the following: (1) The six land use types are ranked by Intersection over Union (IoU) and F1 score as follows: water bodies>buildings>forest and grassland>orchards>cultivated land>other. Excluding the “other” category, the F1 score for all specific land-use types exceeds 0.8, indicating strong applicability and stable performance of the method under complex arid environmental conditions. (2) The “zoning-stratification” strategy significantly enhances extraction accuracy compared to conventional single-model extraction methods. The proposed method achieves a mean IoU (mIoU) of 0.783, a Kappa coefficient of 0.762, and a Macro-F1 of 0.871, which correspond to relative improvements of 17.92%, 9.17%, and 23.90%, respectively, over the baseline method. (3) The four deep learning models used in the study, namely Deeplab v3+, HRNet, REAUNet+UAED, and U-Net3+, each demonstrate optimal performance for their respective land-use types, validating the effectiveness of the stratified object extraction strategy. In summary, this study offers a fresh perspective on integrating geoscientific analytical approaches with advanced deep learning algorithms for fine-grained land-use remote-sensing mapping in arid regions.

Key words: northwestern arid regions, fine-grained land-use remote-sensing mapping, geographical zoning, stratified object extraction, deep learning

LI Jinzhen, WU Tianjun, LUO Jiancheng, HU Xiaodong, ZHOU Zhongfa, CAO Feng. Fine-grained land-use remote-sensing mapping in arid regions using deep learning with “zoning-stratification”[J].Arid Land Geography, 2026, 49(10): 2122-2133.

Figures/Tables 12

Fig. 1

Fig. 2

Fig. 3

Tab. 1

Tab. 2

Fig. 4

Tab. 3

Fig. 5

Tab. 4

Fig. 6

Tab. 5

Tab. 6

References 32

[1]	Gaur S, Mittal A, Bandyopadhyay A, et al. Spatio-temporal analysis of land use and land cover change: A systematic model inter-comparison driven by integrated modelling techniques[J]. International Journal of Remote Sensing, 2020, 41(23): 9229-9255. doi: 10.1080/01431161.2020.1815890
[2]	金璞珺. 哈密市土地利用时空演变对生态系统碳储量的影响研究[D]. 乌鲁木齐: 新疆农业大学, 2023.
	[Jin Pujun. Study on the impact of temporal and spatial evolution of land use on carbon storage of ecosystem in Hami City[D]. Urumqi: Xinjiang Agricultural University, 2023. ]
[3]	宋小龙, 李陇堂, 任婕, 等. 西北干旱生态地理区NDVI时空变化特征及其驱动力[J]. 干旱区地理, 2025, 48(6): 951-962. doi: 10.12118/j.issn.1000-6060.2024.548
	[Song Xiaolong, Li Longtang, Ren Jie, et al. Spatio-temporal variation characteristics and driving factors of NDVI in the arid and semi-arid region of northwest China[J]. Arid Land Geography, 2025, 48(6): 951-962. ] doi: 10.12118/j.issn.1000-6060.2024.548
[4]	Middleton N, Thomas D. World atlas of desertification[M]. London: Arnold, 1997: 1-182.
[5]	张惠怡. 基于机器学习算法的城市土地利用精细制图--以郑州市为例[D]. 开封: 河南大学, 2021.
	[Zhang Huiyi. Detailed mapping urban land use based on machine learning algorithm: A case study of Zhengzhou[D]. Kaifeng: Henan University, 2021. ]
[6]	王金枝, 周广胜, 吕晓敏, 等. 基于光谱特征的拉萨市人造地表遥感提取研究[J]. 光谱学与光谱分析, 2025, 45(4): 1061-1070.
	[Wang Jinzhi, Zhou Guangsheng, Lü Xiaomin, et al. Research on remote sensing extraction of artificial surface in Lhasa City based on spectral features[J]. Spectroscopy and Spectral Analysis, 2025, 45(4): 1061-1070. ]
[7]	李述. 干旱、半干旱区土地利用/覆盖变化与荒漠化的遥感综合研究[D]. 兰州: 兰州大学, 2006.
	[Li Shu. A synthetical study on land use/cover change and desertification in arid and semiarid region[D]. Lanzhou: Lanzhou University, 2006. ]
[8]	刘志飞, 杨雪梅, 王景瑞, 等. 基于高光谱反射特征的荒漠类型可分性评价[J]. 干旱区研究, 2025, 42(1): 141-153. doi: 10.13866/j.azr.2025.01.13
	[Liu Zhifei, Yang Xuemei, Wang Jingrui, et al. Class separability evaluation of desert types based on the hyperspectral reflectance characteristics[J]. Arid Zone Research, 2025, 42(1): 141-153. ] doi: 10.13866/j.azr.2025.01.13
[9]	李爱华. 基于遥感影像认知理解的干旱半干旱地区土地利用/覆盖自动分类方法研究[D]. 兰州: 兰州大学, 2008.
	[Li Aihua. Study on automatic classification for land use/cover in arid and semi-arid area based upon remotely sensed image cognition[D]. Lanzhou: Lanzhou University, 2008. ]
[10]	夏子龙, 杜培军, 郭山川, 等. 呼包鄂榆城市群遥感土地覆被产品评估分析与集成学习[J]. 干旱区地理, 2024, 47(8): 1367-1379. doi: 10.12118/j.issn.1000-6060.2023.718
	[Xia Zilong, Du Peijun, Guo Shanchuan, et al. Evaluation and ensemble learning of remote sensing land-cover products in the Hohhot-Baotou-Ordos-Yulin urban agglomeration[J]. Arid Land Geography, 2024, 47(8): 1367-1379. ] doi: 10.12118/j.issn.1000-6060.2023.718
[11]	Hu W, Huang Y Y, Wei L, et al. Deep convolutional neural networks for hyperspectral image classification[J]. Journal of Sensors, 2015: 258619, doi: 10.1155/2015/258619.
[12]	Shelhamer E, Long J, Darrell T. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(4): 640-651. doi: 10.1109/TPAMI.2016.2572683 pmid: 27244717
[13]	Chen L C, Zhu Y K, Papandreou G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]// FerrariV, HebertM, SminchisescuC, et al. Proceedings of the European Conference on Computer Vision (ECCV). Cham: Springer, 2018: 801-818.
[14]	Zheng S X, Lu J C, Zhao H S, et al. Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers[C]// KletteR, RiveraM, LiZ, et al. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Computer Society Press, 2021: 6881-6890.
[15]	Xie E Z, Wang W H, Yu Z D, et al. SegFormer: Simple and efficient design for semantic segmentation with transformers[J]. Advances in Neural Information Processing Systems, 2021, 34: 12077- 12090.
[16]	周楠. 基于深度学习的精细化土地覆盖变化遥感提取方法研究[D]. 北京: 中国科学院大学(中国科学院空天信息创新研究院), 2022.
	[Zhou Nan. Deep learning-based fine-grained land cover change detection on remote sensing images[D]. Beijing: University of Chinese Academy of Sciences (Aerospace Information Research Institute, Chinese Academy of Sciences), 2022. ]
[17]	赵馨. 深度学习支持下石漠化耕地提取及其生态风险评价--以关岭县为例[D]. 贵阳: 贵州师范大学, 2021.
	[Zhao Xin. Extraction of farmland rocky desertification based on deep learning and its landscape ecological risk assessment: A case study in Guanling County[D]. Guiyang: Guizhou Normal University, 2021. ]
[18]	盛晟, 万芳琦, 林康聆, 等. 基于分层特征提取和多尺度特征融合的高分辨率遥感影像水体提取深度学习算法[J]. 人民珠江, 2024, 45(2): 45-52.
	[Sheng Sheng, Wan Fangqi, Lin Kangling, et al. Deep learning algorithm for water body extraction from high-resolution remote sensing images based on hierarchical feature extraction and multi-scale feature fusion[J]. Pearl River, 2024, 45(2): 45-52. ]
[19]	周登继. 基于深度学习的高分影像城市建筑物提取方法研究[D]. 北京: 中国科学院大学(中国科学院空天信息创新研究院), 2022.
	[Zhou Dengji. Research on extraction method of urban buildings from high resolution remote sensing imagery based on deep learning[D]. Beijing: University of Chinese Academy of Sciences (Aerospace Information Research Institute, Chinese Academy of Sciences), 2022. ]
[20]	侯熠, 许鑫, 齐珂, 等. 复杂遥感场景下的土地利用分类与制图方法研究[J]. 测绘科学, 2025, 50(7): 195-207.
	[Hou Yi, Xu Xin, Qi Ke, et al. Research on land use classification and mapping methods in complex remote sensing scenarios[J]. Science of Surveying and Mapping, 2025, 50(7): 195-207. ]
[21]	王惠林. 基于知识的遥感图像分类方法研究--以腾格里沙漠南部地区为例[D]. 兰州: 兰州大学, 2007.
	[Wang Huilin. Knowledge based remotely sensed imagery classification method research: A case study of the south area of the Tengger Desert[D]. Lanzhou: Lanzhou University, 2007. ]
[22]	赵馨, 周忠发, 王玲玉, 等. 喀斯特山区石漠化耕地遥感精准提取与分析--以贵州省北盘江镇与花江镇为例[J]. 热带地理, 2020, 40(2): 289-302. doi: 10.13284/j.cnki.rddl.003233
	[Zhao Xin, Zhou Zhongfa, Wang Lingyu, et al. Extraction and analysis of cultivated land experiencing rocky desertification in Karst Mountain Areas based on remote sensing: A case study of Beipanjiang Town and Huajiang Town in Guizhou Province[J]. Tropical Geography, 2020, 40(2): 289-302. ] doi: 10.13284/j.cnki.rddl.003233
[23]	Zhang J, Wu T J, Luo J C, et al. Toward agricultural cultivation parcels extraction in the complex mountainous areas using prior information and deep learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 4402414, doi: 10.1109/TGRS.2025.3530615.
[24]	骆剑承, 吴田军, 吴志峰, 等. 遥感大数据智能计算[M]. 北京: 科学出版社, 2020: 1-388.
	[Luo Jiancheng, Wu Tianjun, Wu Zhifeng, et al. Intelligent computation of remote sensing big data[M]. Beijing: Science Press, 2020: 1-388. ]
[25]	张静, 吴田军, 骆剑承, 等. 地理分析与遥感计算耦合的土地利用空间多粒度解构研究[J]. 地球信息科学学报, 2026, 28(3): 573-590. doi: 10.12082/dqxxkx.2025.250098
	[Zhang Jing, Wu Tianjun, Luo Jiancheng, et al. Multi-granularity decomposition of land use space: Integrating geographic analysis and remote sensing computation[J]. Journal of Geo-Information Science, 2026, 28(3): 573-590. ]
[26]	吴田军, 骆剑承, 张新, 等. 基于地理图斑的遥感粒计算与精准应用[J]. 遥感学报, 2023, 27(12): 2774-2795.
	[Wu Tianjun, Luo Jiancheng, Zhang Xin, et al. Remote sensing granular computing and precise applications based on geo-parcels[J]. National Remote Sensing Bulletin, 2023, 27(12): 2774-2795. ] doi: 10.11834/jrs.20211622
[27]	罗桥顺, 党红, 张智光. 基于土地利用/覆盖变化的干旱区生态经济可持续发展研究--以哈密地区为例[C]// 中国测绘学会. 中国测绘学会九届四次理事会暨2008年学术年会论文集. 北京: 测绘出版社, 2008: 615-622.
	[Luo Qiaoshun, Dang Hong, Zhang Zhiguang. Study on the sustainable development of eco-economic in Hami District based on land use/cover change[C]// Chinese Society for Geodesy, Photogrammetry and Cartography. Proceedings of the 4^th Session of the 9^th Council and the 2008 Academic Annual Conference of the Chinese Society for Geodesy, Photogrammetry and Cartography. Beijing: Surveying and Mapping Press, 2008: 615-622. ]
[28]	伊木然江·阿卜来提, 张永福, 孜比布拉·司马义. 基于GlobeLand 30的哈密市2000-2020年土地利用格局变化研究[J]. 水土保持通报, 2021, 41(1): 182-189, 196.
	[Abulaiti Yimuranjiang, Zhang Yongfu, Simayi Zibibula. A study on change of land use pattern in Hami City during 2000-2020 based on GlobeLand 30[J]. Bulletin of Soil and Water Conservation, 2021, 41(1): 182-189, 196. ]
[29]	杨永梅. 区域土地利用结构变化与土地可持续利用研究--以哈密市为例[J]. 中小企业管理与科技(上旬刊), 2010(10): 156.
	[Yang Yongmei. Study on regional changes in land use structure and sustainable land use: A case study in Hami City[J]. Management & Technology of SME, 2010(10): 156. ]
[30]	Wang J D, Sun K, Cheng T H, et al. Deep high-resolution representation learning for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 43(10): 3349-3364. doi: 10.1109/TPAMI.2020.2983686
[31]	袁盼丽, 汪传建, 赵庆展, 等. 基于深度学习的寒旱区多时序影像土地利用及变化监测--以新疆莫索湾垦区为例[J]. 干旱区地理, 2021, 44(6): 1717-1728. doi: 10.12118/j.issn.1000–6060.2021.06.20
	[Yuan Panli, Wang Chuanjian, Zhao Qingzhan, et al. Dynamic monitoring of land-use/land-cover change in cold and arid region based on deep learning: A case study of Mosuowan reclamation area in Xinjiang[J]. Arid Land Geography, 2021, 44(6): 1717-1728. ] doi: 10.12118/j.issn.1000–6060.2021.06.20
[32]	刘巍, 吴志峰, 骆剑承, 等. 深度学习支持下的丘陵山区耕地高分辨率遥感信息分区分层提取方法[J]. 测绘学报, 2021, 50(1): 105-116. doi: 10.11947/j.AGCS.2021.20190448
	[Liu Wei, Wu Zhifeng, Luo Jiancheng, et al. A divided and stratified extraction method of high-resolution remote sensing information for cropland in hilly and mountainous areas based on deep learning[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50(1): 105-116. ] doi: 10.11947/j.AGCS.2021.20190448

地物类型	样本数量/个	图斑数量/个
建筑	17	8305
面状水体	16	40
耕地	79	9636
园地	56	4967
林草地	85	2795
其他	57	398

分层次序	地物类型	匹配模型	提取目标特征
第一层	建筑	Deeplab v3+	城镇建筑：单体占地面积大、形态规则、边界清晰且大规模空间集聚农村建筑：单体占地面积小、小规模团块状或带状分布、蓝白色调屋顶
第二层	面状水体	HRNet	纹理均质、边界清晰
第三层	耕地	REAUNet+UAED	纹理平滑、边界较清晰
第四层	园地	REAUNet+UAED	颗粒状纹理且排列规整，边界模糊
第五层	林草地	U-Net3+	天然牧草地、防护林、城市绿化带和绿化公园等亚类，纹理复杂且边界过渡模糊
第六层	其他	U-Net3+	纹理不均质、植被覆盖度低且边界模糊

地物类型	地块数量/个	地块数量占比/%	总面积/km²	面积占比/%
建筑	26893	58.00	18.76	7.83
面状水体	66	0.14	0.54	0.22
耕地	4846	10.45	47.38	19.76
园地	3227	6.96	15.86	6.62
林草地	5089	10.98	21.11	8.81
其他	6248	13.47	136.06	56.76
总计	46369	100.00	239.71	100.00

地物类型	IoU	UA	PA	F1Score
建筑	0.818	0.924	0.877	0.900
面状水体	0.915	0.979	0.935	0.956
耕地	0.730	0.854	0.820	0.837
园地	0.752	0.868	0.843	0.855
林草地	0.793	0.921	0.852	0.885
其他	0.690	0.819	0.766	0.792

地物类型	精度指标	模型
地物类型	精度指标	Deeplab v3+	HRNet	REAUNet+UAED	U-Net3+
建筑	UA	0.887	0.795	0.734	0.798
	PA	0.839	0.743	0.687	0.763
	F1Score	0.862	0.768	0.710	0.780
面状水体	UA	0.868	0.895	0.801	0.731
	PA	0.811	0.846	0.768	0.697
	F1Score	0.838	0.870	0.784	0.714
耕地	UA	0.821	0.712	0.831	0.669
	PA	0.798	0.754	0.796	0.925
	F1Score	0.809	0.732	0.813	0.777
园地	UA	0.643	0.673	0.822	0.764
	PA	0.929	0.916	0.830	0.705
	F1Score	0.761	0.776	0.826	0.733
林草地	UA	0.741	0.749	0.758	0.820
	PA	0.736	0.779	0.807	0.888
	F1Score	0.738	0.764	0.782	0.853
其他	UA	0.721	0.709	0.664	0.703
	PA	0.750	0.742	0.638	0.861
	F1Score	0.735	0.725	0.650	0.774

Fine-grained land-use remote-sensing mapping in arid regions using deep learning with “zoning-stratification”

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 12

References 32

Related Articles 5

Metrics

Comments

Recommended 0

方法	mIoU	Kappa系数	Macro-F1
单一模型提取	0.664	0.698	0.703
“分区-分层”提取	0.783	0.762	0.871

[1]	AN Jiale, WANG Lei, ZHENG Jianghua, LI Jianhao, ZHAO Guobing, Nigela TUERXUN, FU Rong, LUO Lei. Sentinel-1 and Sentinel-2 based change detection in fire trails under multi-feature combination [J]. Arid Land Geography, 2026, 49(4): 791-803.
[2]	WANG Shengjie, ZHANG Qinghong, SANG Mingjian. Prediction and application of urban PM_2.5 concentration in arid zones based on improved deep learning models [J]. Arid Land Geography, 2026, 49(2): 343-355.
[3]	GUO Jiali, MA Yonggang, PAN Heng, LI Na, SUN Changning, SUN Qian, ZHOU Wenchang, DANG Yuxuan. Comparative analysis of satellite monitoring of soil salinization in Ebinur Lake during spring and summer [J]. Arid Land Geography, 2025, 48(12): 2143-2157.
[4]	YUAN Panli,WANG Chuanjian,ZHAO Qingzhan,WANG Xuewen,REN Yuanyuan,YANG Qiyuan. Dynamic monitoring of land-use/land-cover change in cold and arid region based on deep learning: A case study of Mosuowan reclamation area in Xinjiang [J]. Arid Land Geography, 2021, 44(6): 1717-1728.
[5]	LU Wen-lu, LIU Zhi-feng, HE Chun-yang, XIA Pei. A new method for detecting urban construction land based on Sentinel-1A synthetic aperture radar data and fully convolutional network [J]. Arid Land Geography, 2020, 43(3): 750-760.