基于多时相Sentinel-2影像和SNIC分割算法的优势树种识别

doi:10.11707/j.1001-7488.20220906

摘要/Abstract

摘要：

目的: 探究将简单非迭代聚类超像素分割算法(SNIC)融合到基于多时相数据的树种分类问题中，并对比分析不同时相数据组合对分类结果的影响，实现更高效、更精准的优势树种识别。方法: 以内蒙古旺业甸林场为研究区，在Google Earth Engine(GEE)云计算平台上利用多时相Sentinel-2多光谱数据提取波段反射率特征和光谱指数特征，采用SNIC和支持向量机(SVM)机器学习分类方法，实现面向对象的优势树种识别，并分析不同时相数据组合对优势树种识别精度的影响。结果: 多时相数据组合的分类精度明显高于各季节单时相数据。对比不同多时相数据组合分类结果，春、秋2个季节时间序列组合数据的分类精度与多季节组合数据结果相近，总体精度分别为94.5%、95.0%和95.8%。结论: 基于多时相Sentinel-2影像和SNIC分割算法的面向对象分类方法能够快速、准确识别优势树种，多季节组合数据的分类结果最优，春、秋2个季节时间序列数据也能获得较好分类结果，总体精度与最优结果差距较小。

关键词: 多时相, 简单非迭代聚类超像素分割算法, 树种识别, 时间序列

Abstract:

Objective: The spatial distribution of different tree species is the basis of forest inventory and forest dynamic monitoring, and is of great significance to the protection of forest ecosystems and the sustainable management of forest farms. Method: In this paper, Wangyedian forest farm in Inner Mongolia is selected as the research area. Multi-temporal Sentinel-2 multi-spectral data is used on the Google Earth Engine(GEE)cloud computing platform to extract band reflectance characteristics and spectral index characteristics. The simple non-iterative clustering(SNIC)superpixel segmentation algorithm and the support vector machine(SVM)machine learning classification method are used to identify object-oriented dominant tree species, and the impact of different multi-temporal data combinations on the classification result is analyzed. Result: Experimental result show that the classification accuracy of multi-temporal data combination is significantly higher than that of single-temporal data in each season. Comparing the multi-temporal data combination, The classification accuracy of the combined data of spring and autumn time series is similar to that of multi-season data combination, and their overall accuracy is 94.5%, 95.0%, 95.8%, respectively. Conclusion: The object-oriented classification method proposed in this paper based on multi-temporal data and SNIC algorithm can identify dominant tree species quickly and accurately. Among them, the classification result using multi-season data combination is the best, and the time series data of the spring and autumn seasons can also obtain good classification result, and the overall accuracy is a little lower than the optimal result.

Key words: multi-temporal, simple non-iterative clustering(SNIC), tree species identification, time series

中图分类号:

岳巍,李世明,李增元,刘清旺,庞勇,斯林. 基于多时相Sentinel-2影像和SNIC分割算法的优势树种识别[J]. 林业科学, 2022, 58(9): 60-69.

Wei Yue,Shiming Li,Zengyuan Li,Qingwang Liu,Yong Pang,Lin Si. Identification of Dominant Tree Species Based on Multi-Temporal Sentinel-2 Images and SNIC Segmentation Algorithm[J]. Scientia Silvae Sinicae, 2022, 58(9): 60-69.

图/表 13

表1

图1

表2

图2

表3

图3

图4

图5

图6

表4

图7

图8

图9

参考文献 0

	毕恺艺, 牛铮, 黄妮, 等. 基于Sentinel-2A时序数据和面向对象决策树方法的植被识别. 地理与地理信息科学, 2017, 33 (5): 16- 20.16-20, 27, 127 doi: 10.3969/j.issn.1672-0504.2017.05.003
	Bi K Y , Niu Z , Huang N , et al. Identifying vegetation with decision tree model based on object-oriented method using multi-temporal Sentinel-2A images. Geography and Geo-Information Science, 2017, 33 (5): 16- 20.16-20, 27, 127 doi: 10.3969/j.issn.1672-0504.2017.05.003
	丁世飞, 齐丙娟, 谭红艳. 支持向量机理论与算法研究综述. 电子科技大学学报, 2011, 40 (1): 2- 10.
	Ding S F , Qi B J , Tan H Y . An overview on theory and algorithm of support vector machines. Journal of University of Electronic Science and Technology of China, 2011, 40 (1): 2- 10.
	郭文婷, 张晓丽. 基于Sentinel-2时序多特征的植被分类. 浙江农林大学学报, 2019, 36 (5): 849- 856.
	Guo W T , Zhang X L . Vegetation classification based on a multi-feature Sentinel-2 time series. Journal of Zhejiang A&F University, 2019, 36 (5): 849- 856.
	李军玲, 庞勇, 李增元, 等. 机载AISA Eagle Ⅱ高光谱数据在温带天然林树种分类中的应用. 东北林业大学学报, 2019, 47 (5): 72- 76. doi: 10.3969/j.issn.1000-5382.2019.05.014
	Li J L , Pang Y , Li Z Y , et al. Tree species classification using airborne hyperspectral data in a temperate natural forest. Journal of Northeast Forestry University, 2019, 47 (5): 72- 76. doi: 10.3969/j.issn.1000-5382.2019.05.014
	李哲, 张沁雨, 邱新彩, 等. 基于高分二号遥感影像树种分类的时相及方法选择. 应用生态学报, 2019, 30 (12): 4059- 4070. doi: 10.13287/j.1001-9332.201912.016
	Li Z , Zhang Q Y , Qiu X C , et al. Temporal stage and method selection of tree species classification based on GF-2 remote sensing image. Chinese Journal of Applied Ecology, 2019, 30 (12): 4059- 4070. doi: 10.13287/j.1001-9332.201912.016
	马浩然. 2014. 基于多层次分割的遥感影像面向对象森林分类. 北京: 北京林业大学.
	Ma H R. 2014. Object-based remote sensing image classification of forest based on multi-level segmentation. Beijing: Beijing Forestry University. [in Chinese]
	马倩, 邹焕新, 李美霖, 等. 基于SNIC的双时相SAR图像超像素协同分割算法. 系统工程与电子技术, 2021, 43 (5): 1198- 1209.
	Ma Q , Zou H X , Li M L , et al. Super pixel cooperative segmentation algorithm for bi-temporal SAR image based on SNIC. Systems Engineering and Electronics, 2021, 43 (5): 1198- 1209.
	毛丽君, 李明诗. GEE环境下联合Sentinel主被动遥感数据的国家公园土地覆盖分类. 武汉大学学报(信息科学版), 2021, 1- 19.
	Mao L J , Li M S . Integrating sentinel active and passive data to map land cover in a national park from GEE platform. Geomatics and Information Science of Wuhan University, 2021, 1- 19.
	山东省农业科学院情报资料研究所. 常用农业科技词汇. 济南: 山东科学技术出版社, 1983.
	Institute of Information and Information, Shandong Academy of Agricultural Sciences . Commonly agricultural science and technology vocabulary. Jinan: Shandong Science & Technology Press, 1983.
	谢珠利. 2019. 基于多源高分辨率遥感数据的人工林树种分类研究. 杭州: 浙江农林大学.
	Xie Z L. 2019. Plantation tree species classification with multi-source high resolution remotely sensed data. Hangzhou: Zhejiang A & F University. [in Chinese]
	徐凯健, 田庆久, 徐念旭, 等. 基于时序NDVI与光谱微分变换的森林优势树种识别. 光谱学与光谱分析, 2019, 39 (12): 3794- 3800.
	Xu K J , Tian Q J , Xu N X , et al. Classifying forest dominant trees species based on high dimensional time-series NDVI data and differential transform methods. Spectroscopy and Spectral Analysis, 2019, 39 (12): 3794- 3800.
	尹凌宇, 覃先林, 孙桂芬, 等. 基于高分二号多光谱数据的树种识别方法. 林业资源管理, 2016, (4): 121- 127.
	Yin L Y , Qin X L , Sun G F , et al. Tree species identification method based on GF-2 images. Forest Resources Management, 2016, (4): 121- 127.
	袁旭. 2019. 基于超像素的图像分割方法研究. 武汉: 华中科技大学.
	Yuan X. 2019. Research on image segmentation method based on super-pixel. Wuhan: Huazhong University of Science and Technology. [in Chinese]
	Achanta R , Shaji A , Smith K , et al. SLIC superpixels compared to state-of-the-art superpixel methods. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34 (11): 2274- 2282. doi: 10.1109/TPAMI.2012.120
	Achanta R, Süsstrunk S. 2017. Superpixels and polygons using simple non-iterative clustering. 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA. IEEE, 4895-4904.
	Bolyn C , Michez A , Gaucher P , et al. Forest mapping and species composition using supervised per pixel classification of Sentinel-2 imagery. Biotechnologie, Agronomie Societe et Environnement, 2018, 22 (3): 172- 187.
	Congalton R G . A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment, 1991, 37 (1): 35- 46.
	Deventer V , Ward A D , Gowda P H , et al. Using thematic mapper data to identify contrasting soil plains and tillage practices. Photogrammetric Engineering & Remote Sensing, 1997, 63 (1): 87- 93.
	Drusch M , Del B U , Carlier S , et al. Sentinel-2: ESA's optical high-resolution mission for GMES operational service. Remote Sensing of Environment, 2012, 120, 25- 36.
	Foody G . Assessing the accuracy of remotely sensed data: principles and practices. The Photogrammetric Record, 2010, 25 (130): 204- 205.
	Gao B . NDWI-a normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sensing of Environment, 1996, 58 (3): 257- 266.
	Gitelson A A , Kaufman Y J , Merzlyak M N . Use of a green channel in remote sensing of global vegetation from EOS-MODIS. Remote Sensing of Environment, 1996, 58 (3): 289- 298.
	Hansen M C , Potapov P V , Moore R , et al. High-resolution global maps of 21st-century forest cover change. Science, 2013, 342 (6160): 850- 853.
	Immitzer M , Vuolo F , Atzberger C . First experience with sentinel-2 data for crop and tree species classifications in central Europe. Remote Sensing, 2016, 8 (3): 166.
	Lacaux J P , Tourre Y M , Vignolles C , et al. Classification of ponds from high-spatial resolution remote sensing: application to Rift Valley Fever epidemics in Senegal. Remote Sensing of Environment, 2007, 106 (1): 66- 74.
	Le Maire G L , Francois C , Dufrene E . Towards universal broad leaf chlorophyll indices using PROSPECT simulated database and hyperspectral reflectance measurements. Remote Sensing of Environment, 2004, 89 (1): 1- 28.
	Mahdianpari M , Salehi B , Mohammadimanesh F , et al. Big data for a big country: the first generation of Canadian wetland inventory map at a spatial resolution of 10-m using Sentinel-1 and Sentinel-2 data on the google earth engine cloud computing platform. Canadian Journal of Remote Sensing, 2020, 46 (1): 15- 33.
	Mahdianpari M , Salehi B , Mohammadimanesh F , et al. The first wetland inventory map of newfoundland at a spatial resolution of 10 m using Sentinel-1 and Sentinel-2 data on the google earth engine cloud computing platform. Remote Sensing, 2019, 11 (1): 43.
	Persson M , Lindberg E , Reese H . Tree species classification with multi-temporal sentinel-2 data. Remote Sensing, 2018, 10 (11): 1794.
	Tassi A , Vizzari M . Object-oriented LULC classification in google earth engine combining SNIC, GLCM, and machine learning algorithms. Remote Sensing, 2020, 12 (22): 3776.
	Tsai Y H , Stow D , Chen H L , et al. Mapping vegetation and land use types in Fanjingshan national nature reserve using google earth engine. Remote Sensing, 2018, 10 (6): 927.
	Tucker C J . Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment, 1979, 8 (2): 127- 150.
	Vapnik V . Pattern recognition using generalized portrait method. Automation and remote control, 1963, 24, 774- 780.
	Zhou L , Luo T , Du M Y , et al. Machine learning comparison and parameter setting methods for the detection of dump sites for construction and demolition waste using the google earth engine. Remote Sensing, 2021, 13 (4): 787.

季节Season	序号No.	成像时间Acquisition date	季节Season	序号No.	成像时间Acquisition date
春季Spring	1 2 3 4 5	2019-04-15 2019-05-05 2019-05-10 2020-05-14 2019-05-25	秋季Autumn	8 9 10 11	2019-09-22 2020-09-26 2019-10-07 2020-10-16
夏季Summer	6 7	2019-06-14 2020-08-02	冬季Winter	12	2018-12-16

类别Category	训练样本集Training sample collection	验证样本集Validation sample collection
油松Pinus tabuliformis	68	68
落叶松Larix gmelinii	61	56
其他针叶Other coniferous	15	17
白桦和山杨Betula platyphylla & Populus davidiana	52	47
其他阔叶Other hardwood	34	40
灌木和草地Shrubs and grass	70	70
其他地类Other land-cover	61	63
总数Total	361	361

特征集合 Feature collection	波段Band	中心波长 Central wavelength (S2A/S2B)/nm	带宽Band width (S2A/S2B)/nm
波段反射率 Reflectance of each band	B2(Blue)	497/492	98/98
	B3(Green)	560/559	45/46
	B4(Red)	665/665	38/39
	B5(Rededge1)	704/704	19/20
	B6(Rededge2)	740/739	18/18
	B7(Rededge3)	783/780	28/28
	B8(NIR)	835/833	145/133
	B8A(NIRnarrow)	865/864	33/32
	B11(SWIR1)	1 614/1 610	143/141
	B12(SWIR2)	2 202/2 186	242/238
光谱指数 Spectral indices	光谱指数Spectral index	计算公式Formula	参考文献Reference
	NDVI	$ \frac{{\rm{NIRnarrow-Red}}}{{\rm{NIRnarrow+Red}}}$	(Tucker, 1979)
	NDTI	$ \frac{{\rm{SWIR1-SWIR2}}}{{\rm{SWIR1+SWIR2}}}$	(Deventer et al., 1997)
	NDWI	$ \frac{{\rm{NIRnarrow-SWIR1}}}{{\rm{NIRnarrow+SWIR1}}}$	(Gao, 1996)
	NDVIre	$ \frac{{\rm{NIRnarrow-Rededge1}}}{{\rm{NIRnarrow+Rededge1}}}$	(Gitelson et al., 1996)
	NHI	$ \frac{{\rm{SWIR1-Green}}}{{\rm{SWIR1+Green}}}$	(Lacaux et al., 2007)
	SR_Bre1	$ \frac{{\rm{Blue}}}{{\rm{Rededge1}}}$	(Le Maire et al., 2004)

类别 Category	春季时间序列 Time series of spring		秋季时间序列 Time series of autumn		多季节组合 Multi-season combination
类别 Category	生产者精度 Producer accuracy(%)	用户精度 User accuracy(%)	生产者精度 Producer accuracy(%)	用户精度 User accuracy(%)	生产者精度 Producer accuracy(%)	用户精度 User accuracy(%)
油松Pinus tabuliformis	98.5	95.7	98.5	98.5	98.5	98.5
落叶松Larix gmelinii	94.6	94.6	94.6	91.4	96.4	94.7
其他针叶Other coniferous	82.4	100.0	94.1	100.0	100.0	100.0
白桦和山杨Betula platyphylla & Populus davidiana	100.0	83.9	100.0	87.0	97.9	90.2
其他阔叶Other hardwood	72.5	100.0	77.5	96.9	82.5	97.1
总体精度Overall Accuracy(%)	94.5		95.0		95.8
Kappa系数Kappa coefficient	0.934		0.941		0.951

[1]	黄翀,张晨晨,刘庆生,李贺,杨晓梅,刘高焕. 结合光学与雷达影像多特征的热带典型人工林树种精细识别[J]. 林业科学, 2021, 57(7): 80-91.
[2]	钟莉,陈芸芝,汪小钦. 基于Landsat时序数据的森林干扰监测[J]. 林业科学, 2020, 56(5): 80-88.
[3]	舒琪,胡璇,徐瑞晶,商泽安,漆良华. 海南岛甘什岭青梅种群结构与动态[J]. 林业科学, 2020, 56(5): 160-167.
[4]	宋维,高超,赵玥,赵燕东. 基于LSTM的活立木茎干水分缺失数据填补方法[J]. 林业科学, 2020, 56(2): 134-141.
[5]	栗旭升,李虎,陈冬花,刘玉锋,刘赛赛,刘聪芳,胡国庆. 联合GF-5与GF-6卫星数据的多分类器组合亚热带树种识别[J]. 林业科学, 2020, 56(10): 93-104.
[6]	汤孟平, 沈利芬, 赵赛赛, 仇建习, 徐文兵, 庞春梅, 赵明水, 梅爱君. 基于谱分析的天目山毛竹林生长周期[J]. 林业科学, 2014, 50(9): 184-188.
[7]	王馨爽, 陈尔学, 李增元, 姚顽强, 赵磊. 多时相双极化SAR影像林地类型分类方法[J]. 林业科学, 2014, 50(3): 83-91.
[8]	卢杰, 郭其强, 郑维列, 徐阿生. 藏东南高山松种群结构及动态特征[J]. 林业科学, 2013, 49(8): 154-160.
[9]	刘丽娟;;庞勇;Zhang Xiaoyang;Svein Solberg;范文义;李增元;李明泽. 基于时间序列MODIS EVI数据的森林生长异常监测[J]. 林业科学, 2012, 48(2): 54-62.
[10]	李春明;唐守正. 基于非线性混合模型的落叶松云冷杉林分断面积模型[J]. 林业科学, 2010, 46(7): 106-113.
[11]	王晓慧李增元高志海白黎娜车学俭王琫瑜. 沙化土地信息提取研究[J]. 林业科学, 2005, 41(3): 82-87.
[12]	潘存德. 林木直径分布预测动态模型的研究[J]. , 1990, 26(5): 470-474.