A minimal-interaction framework for accurate and batch extraction of geospatial objects from remote sensing imagery

doi:10.11947/j.AGCS.2025.20250161

Abstract

Abstract:

High-resolution remote sensing image object extraction is a critical technology supporting key areas such as smart city development and natural resource monitoring. However, existing fully automatic methods still face dual challenges in practical applications—limited model adaptability and high manual annotation costs. To address these issues, this paper proposes a high-precision object extraction framework for remote sensing imagery based on minimal interactions (e.g., points, strokes, boxes). By systematically analyzing the limitations of current interactive segmentation techniques, we innovatively construct a unified extraction framework that integrates precise interactive segmentation and batch identical-object detection. The framework comprises two core algorithms: ①A one-shot precision extraction algorithm based on fine-tuning strategies, enabling high-quality object segmentation under minimal interaction; ②A rapid detection algorithm for identical objects, which leverages existing segmentation masks to achieve efficient batch annotation of identical objects. In addition, the extraction framework includes empirical post-processing of geospatial extraction results to obtain vector extraction results. Experimental results on typical facet objects such as buildings, water bodies, and forested areas demonstrate that the proposed method significantly reduces user interactions while maintaining high segmentation accuracy. It outperforms state-of-the-art general-purpose segmentation models, such as segment anything model (SAM) and EISeg. This study provides an innovative solution for efficient annotation of remote sensing image samples and offers significant potential for advancing the automation and practical utility of intelligent remote sensing interpretation.

Key words: graphical prompts, remote sensing imagery, deep learning, interactive segmentation, object extraction

CLC Number:

P237

Zhili ZHANG, Huiwei JIANG, Xiangyun HU. A minimal-interaction framework for accurate and batch extraction of geospatial objects from remote sensing imagery[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(10): 1863-1876.

Figures/Tables 13

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Tab. 1

Fig. 8

Fig. 9

Tab. 2

Fig. 10

Fig. 11

References 23

[1]	张兵, 杨晓梅, 高连如, 等. 遥感大数据智能解译的地理学认知模型与方法[J]. 测绘学报, 2022, 51(7): 1398-1415. DOI: . doi: 10.11947/j.AGCS.2022.20220279
	ZHANG Bing, YANG Xiaomei, GAO Lianru, et al. Geo-cognitive models and methods for intelligent interpretation of remotely sensed big data[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(7): 1398-1415. DOI: . doi: 10.11947/j.AGCS.2022.20220279
[2]	李德仁. 展望大数据时代的地球空间信息学[J]. 测绘学报, 2016, 45(4): 379-384. DOI: . doi: 10.11947/j.AGCS.2016.20160057
	LI Deren. Towards geo-spatial information science in big data era[J]. Acta Geodaetica et Cartographica Sinica, 2016, 45(4): 379-384. DOI: . doi: 10.11947/j.AGCS.2016.20160057
[3]	龚健雅. 人工智能时代测绘遥感技术的发展机遇与挑战[J]. 武汉大学学报(信息科学版), 2018, 43(12): 1788-1796.
	GONG Jianya. Chances and challenges for development of surveying and remote sensing in the age of artificial intelligence[J]. Geomatics and Information Science of Wuhan University, 2018, 43(12): 1788-1796.
[4]	BOYKOV Y Y, JOLLY M P. Interactive graph cuts for optimal boundary & region segmentation of objects in N-D images[C]//Proceedings of the 8th IEEE International Conference on Computer Vision. Vancouver: IEEE Computer Society, 2001: 105-112.
[5]	ROTHER C, KOLMOGOROV V, BLAKE A. GrabCut: interactive foreground extraction using iterated graph cut[J]. ACM Transactions on Graphics, 2004, 23(3): 309-314.
[6]	GRADY L. Random walks for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2006, 28(11): 1768-1783.
[7]	XU Ning, PRICE B, COHEN S, et al. Deep interactive object selection[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 373-381.
[8]	LIEW J, WEI Yunchao, XIONG Wei, et al. Regional interactive image segmentation networks[C]//Proceedings of 2017 IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 2746-2754.
[9]	LEE K M, MYEONG H, SONG G. SeedNet: automatic seed generation with deep reinforcement learning for robust interactive segmentation[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 1760-1768.
[10]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proceedings of the 31st Conference on Neural Information Processing Systems. Long Beach: NIPS, 2017.
[11]	KIRILLOV A, MINTUN E, RAVI N, et al. Segment anything[C]//Proceedings of 2023 IEEE/CVF International Conference on Computer Vision. Paris: IEEE, 2023: 3992-4003.
[12]	OSCO L P, WU Qiusheng, DE LEMOS E L, et al. The segment anything model (SAM) for remote sensing applications: from zero to one shot[J]. International Journal of Applied Earth Observation and Geoinformation, 2023, 124: 103540.
[13]	HAO Y, LIU Y, CHEN Y, et al. Eiseg: an efficient interactive segmentation tool based on paddlepaddle[EB/OL]. [2025-02-08]. https://arxiv.org/pdf/2210.08788.
[14]	WANG Wei. Advanced auto labeling solution with added features[EB/OL]. [2025-02-08]. https://github.com/CVHub520/X-AnyLabeling.
[15]	ZHANG Zhili, HU Xiangyun, YANG Yue, et al. High-quality one-shot interactive segmentation for remote sensing images via hybrid adapter-enhanced foundation models[J]. International Journal of Applied Earth Observation and Geoinformation, 2025, 139: 104466.
[16]	ZHANG Zhili, XU Jiabo, HU Xiangyun, et al. Faster interactive segmentation of identical-class objects with one mask in high-resolution remotely sensed imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 4500516.
[17]	ZHANG Zhili, ZHANG Qi, HU Xiangyun, et al. On the automatic quality assessment of annotated sample data for object extraction from remote sensing imagery[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 201: 153-173.
[18]	BOGUSZEWSKI A, BATORSKI D, ZIEMBA-JANKOWSKA N, et al. LandCover. ai: dataset for automatic mapping of buildings, woodlands, water and roads from aerial imagery[C]//Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Nashville: IEEE, 2021: 1102-1110.
[19]	GLOROT X, BENGIO Y. Understanding the difficulty of training deep feedforward neural networks[C]//Proceedings of the 13th International Conference on Artificial Intelligence and Statistics. Sardinia: [s. n.], 2010.
[20]	HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Delving deep into rectifiers: surpassing human-level performance on ImageNet classification[C]//Proceedings of 2015 IEEE International Conference on Computer Vision. Santiago: IEEE, 2015: 1026-1034.
[21]	KINGMA D P, BA J. Adam: a method for stochastic optimization[C]//Proceedings of 2015 International Conference on Learning Representations. San Diego: ICLR, 2015.
[22]	KE Lei, YE Mingqiao, DANELLJAN M, et al. Segment anything in high quality[J]. Advances in Neural Information Processing Systems, 2023, 36: 29914-29934.
[23]	WEI Zhixiang, CHEN Lin, JIN Yi, et al. Stronger fewer & superior: harnessing vision foundation models for domain generalized semantic segmentation[C]//Proceedings of 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2024.

方法	第1组		第2组		第3组
方法	耗时	IoU/（%）	耗时	IoU/（%）	耗时	IoU/（%）
SAM	14 min 47 s	82.94	12 min 38 s	77.51	11 min 17 s	81.82
EISeg	14 min 36 s	85.45	10 min 46 s	77.44	47 s 10 min	83.88
X-AnyLabeling	16 min 53 s	86.32	11 min 58 s	76.15	9 min 37 s	85.19
Rein-Lora	9 min 37 s	86.59	8 min 23 s	77.52	24 s 7 min	86.91
本文方法	7 min 42 s	86.85	8 min 32 s	77.45	7 min 29 s	87.61

方法	第1组		第2组		第3组
方法	耗时	IoU/（%）	耗时	IoU/（%）	耗时	IoU/（%）
SAM	6 min 38 s	91.96	11 min 45 s	89.78	6 min 37 s	89.55
EISeg	7 min 28 s	91.13	10 min 36 s	84.37	12 s 4 min	86.08
X-AnyLabeling	5 min 32 s	91.95	9 min 52 s	84.37	5 min 25 s	87.16
Rein-Lora	4 min 52 s	93.78	6 min 48 s	90.13	35 s 4 min	91.25
本文方法	4 min 38 s	94.24	6 min 41 s	90.38	5 min 3 s	91.33