基于决策融合的SPOT-6影像土地覆被分类研究

doi:10.12082/dqxxkx.2021.200339

地球信息科学学报 ›› 2021, Vol. 23 ›› Issue (5): 928-937.doi: 10.12082/dqxxkx.2021.200339

基于决策融合的SPOT-6影像土地覆被分类研究

李宏达¹^,²^,³^,⁴(), 高小红¹^,²^,³^,⁴^,^*(), 汤敏¹^,²^,³^,⁴

1.青海师范大学地理科学学院,西宁 810008
2.高原科学与可持续发展研究院,西宁 810008
3.青海省自然地理与环境过程重点实验室,西宁 810008
4.青藏高原地表过程与生态保育教育部重点实验室,西宁 810008

收稿日期:2020-06-30 修回日期:2020-09-19 出版日期:2021-05-25 发布日期:2021-07-25
通讯作者: *高小红（1963— ）,女,陕西白水人,博士,教授,研究方向为遥感信息提取与土地覆被变化。E-mail:xiaohonggao226@163.com
作者简介:李宏达（1995— ）,男,湖北荆门人,硕士生,研究方向为遥感应用与地理空间数据分析。E-mail:2395789679@qq.com
基金资助:
青海省科技厅自然科学基金项目(2016-ZJ-907)

Land Cover Classification for SPOT-6 Image from Decision Fusion Method

LI Hongda¹^,²^,³^,⁴(), GAO Xiaohong¹^,²^,³^,⁴^,^*(), TANG Min¹^,²^,³^,⁴

1. College of Geographical Sciences, Qinghai Normal University, Xining 810008, China
2. Academy of Plateau Science and Sustainability, Xining 810008, China
3. Qinghai Province Key Laboratory of Physical Geography and Environmental Process, Xining 810008, China
4. Ministry of Education Key Laboratory of Tibetan Plateau Land Surface Processes and Ecological Conservation, Xining 810008, China

Received:2020-06-30 Revised:2020-09-19 Online:2021-05-25 Published:2021-07-25
Contact: GAO Xiaohong
Supported by:
Natural Science Fund Project of QingHai Science and Technology Department(2016-ZJ-907)

摘要/Abstract

摘要：

多分类器决策融合方法在提高遥感影像分类的准确性和可靠性方面已表现出了巨大潜力,但这一过程中对所有像元多次分类会产生巨大的时间代价,为改善这一问题,本文提出了主分类器的概念。在青海湟水流域确定2个试验区,对7种常用的分类器进行评估,排除精度较低的3种分类器后,选择支持向量机（Support Vector Machine, SVM）、多层感知机（Multilayer Perceptron, MLP）、随机森林（Random Forest, RF）和梯度提升树（Gradient Boosting Decision Tree, GBDT）4种不同的分类器,建立决策规则共同对SPOT-6影像分类。为提高分类效率,以精度最高的GBDT作为主分类器对影像分类后,仅对结果中可信度不高的像元使用多分类器共同决策。研究结果表明,2个区域内主分类器独立完成分类的像元分别占38.10%和65.26%,错分率为1.57%和2.18%;多分类器共同决策的区域,相比GBDT的分类结果,总体精度分别高出2.49%和3.66%。整体上看,决策融合使2个区域的总体分类精度分别提高了1.18%和1.09%,能够有效减少分类结果中的“椒盐噪声”,精度更加均衡。相比现有的决策融合方法,主分类器的使用在保证分类精度的同时有利于分类效率的提高及分类结果保持良好的一致性。

关键词: 决策融合, 多分类器, 主分类器, 机器学习, GBDT, 土地覆被分类, SPOT-6, 湟水流域

Abstract:

The decision fusion of multi-classifiers has shown great potential in improving the accuracy and reliability of remote sensing image classification. However, multi-classification of all pixels is usually time-consuming. In order to solve this problem, this paper puts forward the concept of master classifier based on existing studies. Firstly, two experimental areas were selected in the Huangshui river basin of Qinghai province: region A representing urban area with serious human activities and complex spectra, and region B representing rural and mountainous area with relatively simple spectra. To obtain a high classification accuracy, seven different commonly used classifiers were selected for decision fusion. Using the testing samples, the classifiers with low accuracy were excluded based on the average accuracy of two regions. Excluded classifies were Naive Bayesian (NB), K-NearestNeighbor (KNN), and Decision Tree (DT). The other four classifiers including Gradient Boosting Decision Tree (GBDT), Random Forest (RF), Support Vector Machine (SVM), and Multilayer Perceptron (MLP) were kept to establish the decision rules for SPOT-6 images classification. Particularly, the GBDT with the highest accuracy among all classifiers was regarded as the master classifier. After the classification by GBDT, the pixels with low confidence were classified again using the other three classifiers, and the decision fusion was made together with the result of GBDT to select the best classification result. The results show that, 38.10% and 65.26% of the pixels in the two regions were classified by the master classifier alone, respectively, and the misclassification rate was 1.57% and 2.18%, respectively. For the regions of decision making using multiple classifiers, the overall classification accuracy was respectively 2.49% and 3.66% higher than that using GBDT. On the whole, the decision fusion improved the overall classification accuracy of the two regions by 1.18% and 1.09% respectively, effectively reduced the "salt and pepper " noise in the results, and achieved a more homogeneous classification accuracy. Compared with the existing decision fusion researches, the use of the master classifier not only ensures the accuracy of classification, but also improves the classification efficiency and maintains a good consistency of the classification results.

Key words: decision fusion, multiple classifiers, the master classifier, machine learning, GBDT, land cover classification, SPOT-6, Huangshui river basin

李宏达, 高小红, 汤敏. 基于决策融合的SPOT-6影像土地覆被分类研究[J]. 地球信息科学学报, 2021, 23(5): 928-937.DOI:10.12082/dqxxkx.2021.200339

LI Hongda, GAO Xiaohong, TANG Min. Land Cover Classification for SPOT-6 Image from Decision Fusion Method[J]. Journal of Geo-information Science, 2021, 23(5): 928-937.DOI:10.12082/dqxxkx.2021.200339

图/表 8

图1

表1

表2

图2

图3

图4

表3

图5

参考文献 29

[1]	García M C, Gutiérrez J G, Riquelme J C. A framework for evaluating land use and land cover classification using convolutional neural networks[J]. Remote Sensing, 2019,11(3):274-296. doi: 10.3390/rs11030274
[2]	黎云云, 畅建霞, 王义民, 等. 基于CA-Markov模型的黄河流域土地利用模拟预测研究[J]. 西北农林科技大学学报(自然科学版), 2020,48(12):1-10.
	[ Li Y Y, Chang J X, Wang Y M, et al. Land use simulation and prediction in the Yellow River Basin based on CA-Markov model[J]. Journal of Northwest A&F University (Natural Science Edition), 2020,48(12):1-10. ]
[3]	金鑫, 陈琼, 金彦香. 土地利用/覆被数据精度对流域水文过程模拟的影响研究[J]. 灌溉排水学报, 2018,37(2):108-115.
	[ Jin X, Chen Q, Jin Y X. Effect of different LULC datasets on hydrologic modeling[J]. Journal of Irrigation and Drainage, 2018,37(2):108-115. ]
[4]	Phan T N, Kappas M. Comparison of random forest, k-nearest neighbor, and support vector machine classifiers for land cover classification using sentinel-2 imagery[J]. Sensors, 2018,18(1):18-37. doi: 10.3390/s18010018
[5]	Vasilakos C, Kavroudakis D, Georganta A. Machine learning classification ensemble of multitemporal sentinel-2 images: The case of a mixed mediterranean ecosystem[J]. Remote Sensing, 2020,12(12):2005-2029. doi: 10.3390/rs12122005
[6]	Mas J F, Flores J J. The application of artificial neural networks to the analysis of remotely sensed data[J]. International Journal of Remote Sensing, 2008,29(3):617-663. doi: 10.1080/01431160701352154
[7]	李玲俐. 数据挖掘中分类算法综述[J]. 重庆师范大学学报(自然科学版), 2011,28(4):44-47.
	[ Li L L. A review on classification algorithms in data mining[J]. Journal of Chongqing Normal University (Natural Science Edition), 2011,28(4):44-47. ]
[8]	Breiman L. Random forests[J]. Machine Learning, 2001,45(1):5-32. doi: 10.1023/A:1010933404324
[9]	Ge G, Shi Z J, Zhu Y J, et al. Land use/cover classification in an arid desert-oasis mosaic landscape of China using remote sensed imagery: performance assessment of four machine learning algorithms[J]. Global Ecology and Conservation, 2020,22:971-983.
[10]	Khatami R, Mountrakis G, Stehman S V. A meta-analysis of remote sensing research on supervised pixel-based land-cover image classification processes: General guidelines for practitioners and future research[J]. Remote Sensing of Environment, 2016,177(5):89-100. doi: 10.1016/j.rse.2016.02.028
[11]	Wang X X, Gao X W, Zhang Y Z, et al. Land-cover classification of coastal wetlands using the RF algorithm for worldview-2 and landsat 8 images[J]. Remote Sensing, 2019,11(16):1927-1948. doi: 10.3390/rs11161927
[12]	谷晓天, 高小红, 马慧娟, 等. 复杂地形区土地利用/土地覆被分类机器学习方法比较研究[J]. 遥感技术与应用, 2019,34(1):57-67.
	[ Gu X T, Gao X H, Ma H J, et al. Comparison of machine learning methods for land use/land cover classification in the complicated terrain regions[J]. Remote Sensing Technology and Application, 2019,34(1):57-67. ]
[13]	边增淦, 王文, 江渊. 黑河流域中游地区作物种植结构的遥感提取[J]. 地球信息科学学报, 2019,21(10):1629-1641. doi: 10.12082/dqxxkx.2019.190183
	[ Bian Z G, Wang W, Jiang Y. Remote sensing of cropping structure in the middle reaches of the Heihe river basin[J]. Journal of Geo-information Science, 2019,21(10):1629-1641. ]
[14]	Löw F, Conrad C, Michel U. Decision fusion and non-parametric classifiers for land use mapping using multi-temporal RapidEye data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015,108:191-204. doi: 10.1016/j.isprsjprs.2015.07.001
[15]	Clinton N, Yu L, Gong P. Geographic stacking: Decision fusion to increase global land cover map accuracy[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015,103(5):57-65. doi: 10.1016/j.isprsjprs.2015.02.010
[16]	Kittler J V, Hatef M, Duin R P W, et al. On combining classifiers[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998,20(3):226-239. doi: 10.1109/34.667881
[17]	陈伟锋, 毛政元, 徐伟铭, 等. 基于Adaboost的高分遥感影像自动变化检测方法[J]. 地球信息科学学报, 2018,20(12):1756-1767. doi: 10.12082/dqxxkx.2018.180353
	[ Chen W F, Mao Z Y, Xu W M, et al. Automatic change detection approach for high-resolution remotely sensed images based on adaboost algorithm[J]. Journal of Geo-information Science, 2018,20(12):1756-1767. ]
[18]	Freund Y, Schapire R E. A decision-theoretic generalization of on-line learning and an application to boosting[J]. Journal of Computer and System Sciences, 1997,55(1):119-139. doi: 10.1006/jcss.1997.1504
[19]	Sun L, Tang L N, Shao G F, et al. A machine learning-based classification system for urban built-up areas using multiple classifiers and data sources[J]. Remote Sensing, 2020,12(1):91-108. doi: 10.3390/rs12010091
[20]	Yang L B, Mansaray L R, Huang J F, et al. Optimal segmentation scale parameter, feature subset and classification algorithm for geographic object-based crop recognition using multisource satellite imagery[J]. Remote Sensing, 2019,11(5):514-536. doi: 10.3390/rs11050514
[21]	马慧娟, 高小红, 谷晓天. 随机森林方法支持的复杂地形区土地利用/土地覆被分类研究[J]. 地球信息科学学报, 2019,21(3):359-371. doi: 10.12082/dpxxkx.2019.180346
	[ Ma H J, Gao X H, Gu X T. Random forest classification of landsat 8 imagery for the complex terrain area based on the combination of spectral, topographic and texture information[J]. Journal of Geo-information Science, 2019,21(3):359-371. ]
[22]	吴宏安, 蒋建军, 张海龙, 等. 比值居民地指数在城镇信息提取中的应用[J]. 南京师大学报(自然科学版), 2006,29(3):118-121.
	[ Wu H, Jiang J J, Zhang H L, et al. Application of ratio resident-area index to retrieve urban residential areas based on landsat TM data[J]. Journal of Nanjing Normal University (Natural Science Edition), 2006,29(3):118-121. ]
[23]	刘畅, 杨康, 程亮, 等. Landsat8不透水面遥感信息提取方法对比[J]. 国土资源遥感, 2019,31(3):148-156.
	[ Liu C, Yang K, Cheng L, et al. Comparison of Landsat8 impervious surface extraction methods[J]. Remote Sensing for Land & Resources, 2019,31(3):148-156. ]
[24]	Polikar R. Essemble based systems in decision making[J]. IEEE Circuits & Systems Magazine, 2006,6(3):21-45.
[25]	Brown K M, Foody G M, Atkinson P M. Estimating per-pixel thematic uncertainty in remote sensing classifications[J]. International Journal of Remote Sensing, 2009,30(1):209-229. doi: 10.1080/01431160802290568
[26]	刘纪远. 国家资源环境遥感宏观调查与动态监测研究[J]. 遥感学报, 1997,1(3):225-230.
	[ Liu J Y. Study on national resources & environment survey and dynamic monitoring using remote sensing[J]. Journal of Remote Sensing, 1997,1(3):225-230. ]
[27]	Nguyen H T T, Doan T M, Tomppo E, et al. Land use/land cover mapping using multitemporal sentinel-2 imagery and four classification methods: A case study from Dak Nong, Vietnam[J]. Remote Sensing, 2020,12(9):1367-1393. doi: 10.3390/rs12091367
[28]	Sothe C, Almeida C M, Veraldo L, et al. Evaluating sentinel-2 and landsat-8 data to map sucessional forest stages in a subtropical forest in southern Brazil[J]. Remote Sensing, 2017,9(8):838-859. doi: 10.3390/rs9080838
[29]	Cui B G, Cui J D, Hao S Y, et al. Spectral-spatial hyperspectral image classification based on superpixel and multi-classifier fusion[J]. International Journal of Remote Sensing, 2020,41(16):6157-6182. doi: 10.1080/01431161.2020.1736730

	区域A		区域B
	训练样本	测试样本	训练样本	测试样本
耕地	300	139	174	84
林地	264	85	94	52
草地	391	107	241	90
水域	107	57	36	25
城乡建设用地	415	113	131	66
未利用土地	257	66	100	18

	GBDT	SVM	RF	MLP	DT	KNN	NB
区域A /%	88.16	88.01	87.01	86.74	80.66	76.14	75.35
区域B /%	90.53	88.61	89.35	87.24	85.39	79.47	70.93
平均精度/%	89.35	88.31	88.18	86.99	83.03	77.81	73.14

	区域A					区域B
	GBDT		决策融合		GBDT		决策融合
	PA/%	UA/%	PA/%	UA/%	PA/%		UA/%	PA/%	UA/%
耕地	91.78	88.16	91.78	87.01	86.64	91.85	83.40	91.56
林地	90.00	98.44	94.29	98.51	96.62	98.28	98.31	98.31
草地	84.88	76.04	88.37	79.17	88.84	80.83	92.15	82.29
水域	93.68	98.33	92.06	98.31	96.77	97.40	97.42	99.34
城乡建设用地	87.50	80.46	86.25	83.13	83.39	83.09	86.72	87.36
未利用土地	83.33	95.89	84.52	95.95	96.52	99.49	97.51	98.49
总体精度/%	88.16		89.25		90.76		91.94
Kappa系数	0.86		0.87		0.89		0.90

基于决策融合的SPOT-6影像土地覆被分类研究

Land Cover Classification for SPOT-6 Image from Decision Fusion Method

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 29

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	戚梦, 陈楠, 林偲蔚, 周千千. 引入集水区复杂网络的中国地貌识别研究[J]. 地球信息科学学报, 2023, 25(5): 909-923.
[2]	李琳叶, 李艳艳, 陈传法, 刘妍, 刘雅婷, 刘盼盼. 林区数字高程模型修正方法：顾及高程自相关的后向传播神经网络模型[J]. 地球信息科学学报, 2023, 25(5): 935-952.
[3]	柯日宏, 吴升, 柯玮文. 一种识别共享单车潮汐点的时空模型和基于KNN-LightGBM的租还需求预测方法[J]. 地球信息科学学报, 2023, 25(4): 741-753.
[4]	李心雨, 闫浩文, 王卓, 王炳瑄. 街景图像与机器学习相结合的道路环境安全感知评价与影响因素分析[J]. 地球信息科学学报, 2023, 25(4): 852-865.
[5]	黄帅元, 董有福, 李海鹏. 黄土高原区SRTM1 DEM高程误差校正模型构建及对比分析[J]. 地球信息科学学报, 2023, 25(3): 669-681.
[6]	谢晓苇, 李代超, 卢嘉奇, 吴升, 许芳年. 基于移动监测数据的不同城市场景下PM_2.5浓度精细模拟与时空特征解析[J]. 地球信息科学学报, 2022, 24(8): 1459-1474.
[7]	黄钦, 杨波, 徐新创, 郝汉舟, 梁莉莉, 王敏. 基于多源空间数据和随机森林模型的长沙市茶颜悦色门店选址与预测研究[J]. 地球信息科学学报, 2022, 24(4): 723-737.
[8]	刘智谦, 吕建军, 姚尧, 张嘉琪, 寇世浩, 关庆锋. 基于街景图像的可解释性城市感知模型研究方法[J]. 地球信息科学学报, 2022, 24(10): 2045-2057.
[9]	邢晓语, 杨秀春, 徐斌, 金云翔, 郭剑, 陈昂, 杨东, 王平, 朱立博. 基于随机森林算法的草原地上生物量遥感估算方法研究[J]. 地球信息科学学报, 2021, 23(7): 1312-1324.
[10]	邓雅文, 凌子燕, 孙娜, 吕金霞. 基于广义回归神经网络的京津冀地区土壤湿度遥感逐日估算研究[J]. 地球信息科学学报, 2021, 23(4): 749-761.
[11]	尹凌, 刘康, 张浩, 奚桂锴, 李璇, 李子垠, 薛建章. 耦合人群移动的COVID-19传染病模型研究进展[J]. 地球信息科学学报, 2021, 23(11): 1894-1909.
[12]	刘明杰, 徐卓揆, 郜允兵, 杨晶, 潘瑜春, 高秉博, 周艳兵, 周万鹏, 王凌. 基于机器学习的稀疏样本下的土壤有机质估算方法[J]. 地球信息科学学报, 2020, 22(9): 1799-1813.
[13]	陈昂, 杨秀春, 徐斌, 金云翔, 张文博, 郭剑, 邢晓语, 杨东. 基于面向对象与深度学习的榆树疏林识别方法研究[J]. 地球信息科学学报, 2020, 22(9): 1897-1909.
[14]	刘新, 赵宁, 郭金运, 郭斌. 基于LSTM神经网络的青藏高原月降水量预测[J]. 地球信息科学学报, 2020, 22(8): 1617-1629.
[15]	李婉, 牛陆, 陈虹, 吴骅. 基于随机森林算法的地表温度鲁棒降尺度方法[J]. 地球信息科学学报, 2020, 22(8): 1666-1678.