EN中文

Journal of Geo-information Science >

2018 , Vol. 20 >Issue 3: 396 - 404

DOI: https://doi.org/10.12082/dqxxkx.2018.170553

Study on Extraction of Citrus Orchard in Gannan Region Based on Google Earth Engine Platform

  • XU Hanzeyu , 1 ,
  • LIU Chong 1 ,
  • WANG Junbang 2 ,
  • QI Shuhua , 1, *
Expand
  • 1. School of Geography and Environment, Key Laboratory of Poyang Lake Wetland and Watershed Research, Ministry of Education, Jiangxi Normal University, Nanchang 330022, China
  • 2. Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
*Corresponding author: QI Shuhua, E-mail:

Received date: 2017-11-21

  Request revised date: 2017-12-23

  Online published: 2018-03-20

Supported by

National Natural Science Foundation of China, No.41261069, 41601453

The Collaborative Innovation Center For Major Ecological Security Issues of Jiangxi Province and Monitoring Implementation, No. JXS-EW-00).

Copyright

《地球信息科学学报》编辑部 所有
Fold

Abstract

Gannan region is located in the southern Jiangxi Province, China. Gannan region includes 2 districts and 15 counties in Ganzhou. It has hilly land resources and its climate conditions are benefit to plant citrus. With the support and guidance from local government, Gannan region has experienced the boom of citrus planting and become the largest citrus production region in China over the past decades. Despite the economic success, the rapid and extensive citrus orchard expansion has brought great concern about ecological consequences. It is necessary to map citrus orchard for understanding the effects of citrus expansion. The objective of this study is to map the citrus orchard in 1990, 1995, 2000, 2005, 2010 and 2016 in Gannan Region. An image composite method was applied and total 2140 tiles of Landsat historical images were employed to generate seasonal images with lowest cloud composite at the pixel level. Random Forest classifier was used to classify multiple dimensional features from spectral, spatial and topographic domains. The image composite and classification were implemented with Google Earth Engine (GEE) platform. Results showed that: (1) GEE can effectively execute complex workflows of remote sensing data processing and information digging. (2) Lowest cloud composite at the pixel level is a reliable method of producing clear seasonal images for the region influenced by cloud and rain frequently. (3) Random forest classifier was suitable for mapping citrus orchard with an average overall accuracy (OA) of 93.15% and a Kappa coefficient of 0.90. (4) The citrus orchard has expanded extensively with the area from 9.77 km2 in 1990 to 2200.34 km2 in 2016. Citrus orchard was becoming clustered especially in Xunwu, Xinfeng and Anyuan and was mainly converted from forest, bush and cropland.

Key words： Gannan region; citrus orchard; Google Earth Engine; lowest cloud coverage; image composite; classification

Cite this article

XU Hanzeyu , LIU Chong , WANG Junbang , QI Shuhua . Study on Extraction of Citrus Orchard in Gannan Region Based on Google Earth Engine Platform[J]. Journal of Geo-information Science, 2018 , 20(3) : 396 -404 . DOI: 10.12082/dqxxkx.2018.170553

1 引言

柑橘是橘、柑、橙、柚、枳等芸香科果树的总称,是世界上主要的经济作物之一,因其可食性高、易贮藏、市场广阔等特点被广泛种植。2008年中国超越巴西成为世界上最大的柑橘产地[1]。联合国粮农组织（FAO）的统计数据显示,2016年中国柑橘种植面积已达5057 km²[2]。种植区主要分布于江西、浙江、福建、贵州等红壤丘陵区,并形成了赣南、湘南和桂北等柑橘优势产区。在赣南地区,已形成了以寻乌、安远、信丰等县为代表的主要柑橘种植区,寻乌县柑橘种植面积最大,章贡区柑橘种植面积最少。
20世纪80年代以来,赣州市委市政府提出“兴国富民”等战略,柑橘种植得到了大规模发展,赣南地区成为我国柑橘重要主产区,有力促进了地方经济增长。同时,由于开发方式粗犷,果园扩张造成的景观结构与生物多样性单一、水土流失加剧等生态问题逐渐引起关注[3]。利用卫星遥感技术开展赣南柑橘果园的扩张动态监测,有利于客观评价柑橘果园扩张的生态影响,也有利于柑橘果园的立体化改造和宏观管理。
随着卫星遥感技术和遥感图像自动化识别算法的发展,遥感技术已被广泛应用于农业生产活动的宏观监测[4,5]。针对广泛种植于发展中国家的棕榈[6]、橡胶[7,8,9]、茶[10]、竹[11,12]和水果[13,14,15,16]等经济作物的遥感监测也开展了大量研究。部分经济作物的遥感提取研究并未利用物候特征的季节变化[17,18]。随着支持向量机（SVM）、随机森林（Random Forest）等机器学习分类方法[19]的出现,多维分类特征参与分类成为可能,但庞大的数据量往往超越了本地单机的处理能力。Google Earth Engine（GEE）平台拥有大量的历史遥感影像数据存档,具有支持并行云端运算的特点,为大规模遥感数据处理与数据挖掘提供了技术平台[20,21]。
本文通过GEE平台,使用多时相Landsat系列影像数据,利用像元级最小云量影像合成方法构建目标年份季节合成影像,使用随机森林分类算法提取赣南地区1990、1995、2000、2005、2010和2016年的柑橘果园。

2 研究区概况

“赣南”特指江西省赣州市所辖的2区15县（图1）。赣州市位于赣江中上游地区,介于北纬 24°29′~27°09′,东经113°54′~116°38′之间。东、南、西侧分别与福建、广东、湖南相邻,总面积约3.94 万km2,占江西省总面积的23.6%。
Fig. 1 The image of the study area

图1 研究区概况

赣州市属亚热带湿润性气候,气候温和,热量丰富,雨量充沛,无霜期长,年均气温18.9 ℃,年平均无霜期287 d,年均降雨量1605 mm,年日照时数1813 h,昼夜温差大,雨热同季。区内以红壤为主,多为山地、丘陵地形,山地占全市面积近60%,丘陵占近20%,境内水系发达。具有开展柑橘种植的优越地貌、土壤与气候条件。
赣州市辖区的寻乌、安远、会昌等11个县被列为国家扶贫开发工作重点县。20世纪70年代以来,赣州市委市政府提出的“兴果富民”、“建设世界著名脐橙主产区”、“培植超百亿元产业集群”等政策,有力推动了赣南柑橘产业的快速发展,较好促进了扶农扶贫工作的展开。柑橘果业已发展成为当地农业主导产业。

3 数据源与研究方法

3.1 遥感数据

（1）Landsat影像
Landsat是美国航天局（National Aeronautics and Space Administration, NASA）的陆地卫星计划,自1972年7月23日起,已发射8颗。Landsat 5于1984 年 3 月发射升空、2013年6月退役,搭载有4波段光机扫描仪MSS（Multi Spectral Scanner）和包含7个波段的多光谱扫描仪 TM（Thematic Mapper）。Landsat 8于2013年2月11日发射升空,携带有包含9个波段的OLI陆地成像仪和TIRS热红外传感器。空间分辨率30 m,时间分辨率16 d。本文所使用的2140景Landsat TM/OLI影像数据来自于美国地质调查局（United States Geological Survey, USGS）网站（https://espa.cr.usgs.gov）与Google Earth Engine（GEE）平台 （https://earthengine.google.com/）,其中1990、1995、2000、2005和2010年分别使用了242、395、415、424、372景Landsat-5 TM影像,共计1848景;2016年使用Landsat-8 OLI影像,共计292景,有关信息和各季节所用的影像景数,如图2所示。
Fig. 2 The Landsat images used to produce seasonal clear images with the lowest cloud composite

图2 最小云量合成所使用的Landsat原始影像数量

（2）SRTM DEM 数据
SRTM（Shuttle Radar Topography Mission）[22]为美国“奋进”号航天飞机搭载SRTM系统进行采集制作的数字高程模型。本文使用SRTMGL1_003数据,空间分辨率30 m。

3.2 样本数据

将研究区主要地物划分为裸地、柑橘果园、耕地、林地、灌丛、城镇建设用地和水域,这7种地物在Google Earth平台提供的高分辨率影像上具有明显的辨识特征。各年份样本选取的流程为：① 通过Google Earth平台提供的2016年前后的高分辨率影像,利用目视判别的方法得到所有地物的样本数据,并加入手持GPS收集的56个柑橘果园样本点,建立2016年样本数据,同时将其作为推导其他年份样本数据的标准集;② 通过对比2016、2015年和各目标年份前后2年,共5年的Landsat影像,从2016年向前,对目标年份样本点的实际地物属性进行判别;③ 将判别、修改得到的目标年份样本数据转为 KML 格式,再次在 Google Earth中根据研究区地物变化的先验知识进行检查（如,开垦于2000年的“柑橘果园”,在不受虫害影响下,基本会一直处于“柑橘果园”的状态;1995年呈现“裸地”特点的稀土矿区,在2016年也会具有“裸地”的特点）。通过上述方法收集得到2010、2005、2000、1995、1990年的样本数据。在缺少长时间序列较高分辨率影像和地面记录的情况下,通过时间序列 Landsat 影像进行样本选取是较为通用的方法[23]。最后将各年份的样本数据按4:1的比例进行随机分配,80%样本用于分类器训练,20%样本用于精度评价（表1）。
Tab. 1 Number of samples for each year

表1 分类样本数量表（个）

年份 裸地 柑橘果园 耕地 林地 灌丛 城镇建设用地 水域 合计
1990 86 93 1136 1866 314 107 150 3752
1995 97 158 1136 1858 314 118 150 3831
2000 100 230 1136 1843 314 126 151 3900
2005 108 351 1131 1823 314 144 152 4023
2010 118 427 1112 1800 314 178 153 4102
2016 120 500 1105 1796 314 190 153 4178

3.3 Google Earth Engine平台

由谷歌、卡内基梅隆大学和美国地质调查局联合开发的GEE是目前世界上先进的PB级地理数据科学分析及可视化平台。GEE面向用户提供海量卫星影像数据集与地理数据集,包括多年历史卫星影像数据（如MODIS、Landsat等）与欧空局的卫星影像数据（如Sentinel等）,同时,GEE提供基于JavaScript和Python语言的API接口、分析算法与工具,方便用户实现大型数据的处理分析与信息挖掘。

3.4 柑橘果园遥感提取

3.4.1 多年份季节性最小云量影像合成
赣南地区多云多雨的气候特征,往往难以获得无云或低云的影像。为充分利用研究区影像的季节信息并克服多云雨影响,采用像元级最小云量影像合成方法,以获取不同季节的干净影像,反映不同地物的物候特征。选择目标年份及其前后各1年共3年相同季节的Landsat数据,在GEE平台中使用Landsat云量计算算法对输入的符合时间和空间范围的Landsat原始影像集进行计算,得到输入数据集的大气表观反射率数据（Top of Atmosphere Reflectance, TOA）和每个像元的云量得分,以云量得分最低的像元重构目标年份该季节最小云量合成影像。
3.4.2 分类特征
利用光谱特征、纹理特征和地形特征参与柑橘果园的遥感提取。其中,光谱特征包括目标年份季节最小云量合成影像的多光谱波段及光谱指数,包括：归一化植被指数（Normalized Difference Vegetation Index, NDVI）[24]、归一化湿度指数（Normalized Difference Moisture Index, NDMI）[25]和调整土壤亮度的植被指数（Soil-Adjusted Vegetation Index, SAVI）[26];通过灰度共生矩阵（Grey-Level Co-Occurrence Matrix GLCM）[27]计算影像各多光谱波段的对比度（contrast）、熵（entropy）和二阶矩（second moment）作为纹理特征;地形特征包括DEM和坡度。根据最小云量合成影像TOA数据[28]分别计算得到光谱特征、纹理特征集以及地形特征,作为各目标年份分类特征集。
3.4.3 随机森林算法
本文使用随机森林（Random Forest,RF）分类算法对各目标年份的输入特征集进行分类（图3）。RF是一种包含多个决策树（Classification And Regression Tree, CART）的机器学习算法[29],已在土地利用变化制图等诸多领域得到广泛应用,并被认为针对多维分类特征具有较为稳健的分类效果[30,31]。研究中将待生成决策树数目设值为500,分裂节点数设置为全部特征平方根。
Fig. 3 Workflow of this study

图3 技术路线图

3.4.4 精度评价
使用6个独立的验证样本分别计算6个目标年份分类结果的总体精度（Overall Accuracy, OA）、KAPPA系数、各地物的最小精度（Minimum Accuracy, MA）[32]等对柑橘果园遥感提取精度进行评价。其中,最小精度为各地物用户精度（User Accuracy, UA）或生产者精度（Producer Accuracy, PA）的最小值。

4 结果与讨论

4.1 柑橘果园扩张动态

1990、1995、2000、2005、2010和2016年柑橘果园分布图（图4）表明,赣南地区柑橘果园扩张迅速,主要集中分布于东北部、东南及南部地区;柑橘果园由1990年的零星分布逐渐形成连片化的聚集分布特征;果园面积由1990年9.77 km2扩大为2016年的2200.34 km2（图5）,李自茂等[33]在《赣南脐橙产业发展报告（2013）》中的数据显示,2013年赣州市柑橘种植面积约1658.03 km2;从分县（区）果园种植情况看,寻乌、安远、信丰3县为最早开始柑橘种植的主要地区,1995年已初具种植规模,随后,寻乌、安远、会昌等县逐渐发展为柑橘主产县,2005年以来,各县果园种植范围呈现大幅扩张趋势。柑橘果园用地主要来源于林地、灌丛和耕地。对比1990年和2016年分类结果得到,约有87.23%、11.08%和1.54%的2016年柑橘果园用地分别来源于1990年时的林地、灌丛和耕地。
Fig. 4 The spatial pattern of citrus orchard in Gannan from 1990 to 2016

图4 1990-2016年典型年份赣南地区柑橘果园分布格局

Fig. 5 Variation of citrus orchard areas for each county in Gannan

图5 赣南地区各县（区）柑橘果园面积变化情况

4.2 柑橘果园遥感提取的精度评价

利用OA、KAPPA系数和最小精度等分类精度评价参数对不同年份柑橘果园提取结果进行评价（图6）,结果表明：各年份的分类结果具有较好的分类精度,平均OA和KAPPA系数分别为93.15%和0.90。OA和KAPPA系数的最大值出现在1990年,最小值出现在2016年,OA与KAPPA标准偏差分别为1.78%和2.16%,各年份的分类精度相对稳定。从各分类地物的最小精度（MA）看,水域的识别精度最高（平均MA 98.41%）;林地、耕地次之（平均MA分别为96.42%和95.31%）;裸地和柑橘果园的平均MA分别为70.94%和66.65%,其中柑橘果园的MA的最小值出现在1995年,最大值出现在2005年。稀疏林地和稀疏灌丛同低龄果园具有较为相似的光谱特征和纹理特征,易产生混淆。误分现象主要出现在定南、兴国、赣县、崇义县等县区;漏分现象主要出现在信丰县。
Fig. 6 Accuracy assessment of classification results

图6 精度评价

5 结论

本文通过GEE平台,利用大量Landsat历史影像得到目标年份各季节最小云量合成影像集,构建包括光谱、纹理、地形特征在内的分类特征集,使用RF分类算法,分别提取了1990-2016年每5年时间间隔的赣南地区柑橘果园空间分布情况,得到以下结论：
（1）GEE平台可以有效应用于大量遥感影像数据的处理及信息挖掘工作;
（2）像元级最小云量影像合成重构影像能够有效克服多云多雨的影响,有利于多云多雨地区地物的遥感识别;
（3）由于稀疏林地和稀疏灌丛同低龄果园具有较为相似的光谱特征和纹理特征,易产生混淆,导致柑橘果园的平均用户和生产者精度分别为85.61%和66.98%,误分现象主要出现在定南、兴国、赣县、崇义县等县区,漏分现象主要出现在信丰县;
（4）赣南地区柑橘果园由1990年9.77 km2迅速扩张为2016年的2200.34 km2,果园分布由1990年的零星分布,逐渐形成规模化、连片化的分布特征,果园用地的主要来源为林地、灌丛和耕地。
柑橘果园中果树的排列具有固定排列模式,在卫星遥感影像上形成比较一致的纹理特征,特别是在高分辨率卫星影像中的纹理特征更为明显,使用较高分辨率和高光谱遥感影像,以像元分割技术和面向对象的分类方法,有望更为准确地识别柑橘果园分布范围。

The authors have declared that no competing interests exist.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Wang Y, Wang Y, Wang L, et al.An ontology-based approach to integration of hilly citrus production knowledge[J]. Computers and Electronics in Agriculture, 2015,113:24-43.Citrus planting is subject to a variety of complex factors, such as terrain and growth-stage. There are a few differences between citrus planting on hills and in other places. In this paper, we focus on the integration of hilly citrus production knowledge. We organize and convert the citrus production knowledge in the forms of texts, tables and pictures in technical reports and books into a hilly citrus fertilization and irrigation ontology. Three citrus decision services (fertilization, nutrient imbalance, and irrigation/drainage) are developed for fruit farmers in Chongqing, China, based on the ontology. We compare the system results and the reference values recommended by citrus experts to ensure that our system can provide correct and helpful advice for growers. The results of the fertilization service are completely consistent with the experts’ expectations. The nutrient imbalance service meets growers’ requests with a 98% accuracy, and the irrigation/drainage service can achieve 94% accuracy. Our work can also be further connected with other agricultural ontologies and integrated to be subclasses of AOS ontology.

DOI

[2]
FAO. Food and Agriculture Organization[EB/OL].[2017-1106]..

[3]
孙永明,叶川,王学雄,等.赣南脐橙果园水土流失现状调查分析[J].水土保持研究,2014,21(2):67-71.为全面摸清赣南脐橙果园水土流失状况,为政府部门宏观决策提供科学依据,采取实地调查与统计分析相结合的方法,就赣南脐橙果园水土流失分布、程度展开调查,结果表明:赣南脐橙果园水土流失总面积10.97万hm2,占果园总面积的92％;流失强度其中轻度流失占53.92％,中度流失占33.62％,强度流失占12.46％,以轻度、中度流失为主、且主要发生在3～7 a及7a以上的果园,1～3 a果园水土流失以强度为主;流失量高达360.20万t,流失区域根据平均侵蚀模数划分了5类,信丰属于平均侵蚀模数最轻区域,崇义县属于侵蚀模数最重区域.

[Sun Y M, Ye C, Wang X X, et al.Investigation and analysis on the present situation of soil erosion in Gannan navel orange orchard[J]. Research of Soil and Water Conservation, 2014,21(2):67-71.]

[4]
Zhong L, Yu L, Li X, et al. Rapid corn and soybean mapping in US Corn Belt and neighboring areas[J]. Scientific Reports, 2016,6(3):32-40).Abstract The goal of this study was to promptly map the extent of corn and soybeans early in the growing season. A classification experiment was conducted for the US Corn Belt and neighboring states, which is the most important production area of corn and soybeans in the world. To improve the timeliness of the classification algorithm, training was completely based on reference data and images from other years, circumventing the need to finish reference data collection in the current season. To account for interannual variability in crop development in the cross-year classification scenario, several innovative strategies were used. A random forest classifier was used in all tests, and MODIS surface reflectance products from the years 2008-2014 were used for training and cross-year validation. It is concluded that the fuzzy classification approach is necessary to achieve satisfactory results with R-squared ~0.9 (compared with the USDA Cropland Data Layer). The year of training data is an important factor, and it is recommended to select a year with similar crop phenology as the mapping year. With this phenology-based and cross-year-training method, in 2015 we mapped the cropping proportion of corn and soybeans around mid-August, when the two crops just reached peak growth.

DOI PMID

[5]
史舟,梁宗正,杨媛媛,等.农业遥感研究现状与展望[J].农业机械学报,2015,46(2):247-260.

[Shi Z, Liang Z Z, Yang Y Y, et al.Status and prospect of agricultural remote sensing[J]. Transactions of the Chinese Society for Agricultural machinery, 2015,46(2):247-260.]

[6]
Cheng Y, Yu L, Cracknell A P, et al.Oil palm mapping using Landsat and PALSAR: A case study in Malaysia[J]. International Journal of Remote Sensing, 2016,37(22):5431-5442.Irrespective of the positive economic benefit or negative environmental impact of the rapid expansion of oil palm plantations in tropical regions, it is important to be able to create accurate land-cover maps for such areas. Optical remote sensing is vulnerable to the effects of clouds, which can limit data availability for the oil palm plantation areas in the humid tropics. The satellite-flown PALSAR (Phased Array type L-band Synthetic Aperture Radar) instrument, which provides all-day/all-weather Earth observations, offers the opportunity to identify and map oil palm plantations in cloudy regions. This study used a Support Vector Machine (SVM) classifier and a Mahalanobis distance (MD) classifier to undertake supervised classifications of Landsat, PALSAR, and combined Landsat and PALSAR data (Landsat+PALSAR) for two locations in peninsular Malaysia. Results indicate that accuracies from Landsat+PALSAR are better than accuracies from Landsat and PALSAR along for both study areas using both classifiers. The extents of the oil palm areas estimated from these maps were compared with values obtained through human photointerpretation of Google Earth images in previous studies. Based on the R2 statistics, it was established that the Landsat+PALSAR combination performed best for both study areas and demonstrated good potential for oil palm plantation mapping.

DOI

[7]
梁守真,陈劲松,吴炳方,等.应用面向对象的决策树模型提取橡胶林信息[J].遥感学报,2015,19(3):485-494.橡胶林的无序和不合理种植引发了一系列的生态问题,快速监测橡胶 林空间分布及动态变化,对橡胶的合理种植、区域生态环境保护以及有关部门的规划决策有重要的指导意义.以MODIS归一化植被指数NDVI时间序列数据和 多时相的Landsat TM数据为基础分析橡胶林的季相和光谱特征,确定橡胶识别的关键时期和特征参数,构建面向对象的决策树分类模型,开展橡胶信息提取研究.结果表明,多时相 的遥感数据可反映橡胶的季相特征,以TM数据为基础计算得到的陆表水分指数LSWI和归一化植被指数NDVI可作为橡胶识别的光谱特征参数,橡胶休眠期是 利用遥感方法进行橡胶提取的最佳时期.相比于单时相数据,利用包含橡胶关键物候期的多时相遥感数据能得到更高的橡胶林提取精度.

DOI

[Liang S Z, Chen J S, Wu B F, et al.Extracting rubber plantation with decision tree model based on an object-oriented method[J]. Journal of Remote Sensing, 2015,19(3):485-494.]

[8]
封志明,刘晓娜,姜鲁光,等.中老缅交界地区橡胶种植的时空格局及其地形因素分析[J].地理学报,2013,68(10):1432-1446.

[Feng Z M, Liu X N, Jiang L G, et al.Spatial-temporal analysis of rubber plantation and its relationship with topographical factors in the border region of China, Laos and Myanmar[J]. Acta Geographica Sinica, 2013,68(10):1432-1446.]

[9]
Beckschäfer P.Obtaining rubber plantation age information from very dense Landsat TM & ETM+ time series data and pixel-based image compositing[J]. Remote Sensing of Environment, 2017,196:89-100.

DOI

[10]
徐伟燕,孙睿,金志凤.基于资源三号卫星影像的茶树种植区提取[J].农业工程学报,2016,32(S1):161-168.

[Xu W Y, Sun R, Jin Z F.Extracting tea plantations based on ZY-3 satellite data[J]. Transactions of the Chinese Society of Agricultural Engineering, 2016,32(S1):161-168.]

[11]
Li M, Li C, Jiang H, et al.Tracking bamboo dynamics in Zhejiang, China, using time-series of Landsat data from 1990 to 2014[J]. International Journal of Remote Sensing, 2016,37(7):1714-1729.Bamboo is an important vegetation type and provides a number of critical ecosystem services. Reliable and consistent information on bamboo distribution is required to better estimate its effect on climate change mitigation and socio-economic development. However, such information is rare over a large spatial area. In this study, we evaluate the contribution of different features in the identification of bamboo stands and determine a more discriminative set of features. We propose a bamboo mapping system including feature extraction and feature selection and derive the long-term trends of bamboo distribution in Zhejiang Province, China, using time-series of Landsat data from 1990 to 2014, with an increment of 502years (1990, 1995, 2000, 2005, 2010, and 2014). The resultant maps of bamboo in the six epochs were evaluated using independent validation samples. The overall accuracies (OAs) of all six epochs range from 85.9% to 90.7%. We found that bamboo distribution in Zhejiang substantially increased from 1990 to 2014, particularly during the 2000s. Based on the produced maps, the area of bamboo in this region increased from 536302±0249002km2in 1990 to 1167102±0265302km2in 2014, which is consistent with the National Forest Resource Inventory (NFRI) data. Our study demonstrates the capability of time-series of Landsat data for continuous monitoring of bamboo at a large spatial scale. 08 2016 Informa UK Limited, trading as Taylor & Francis Group.

DOI

[12]
Liu C, Xiong T, Gong P, et al.Improving large-scale moso bamboo mapping based on dense Landsat time series and auxiliary data: A case study in Fujian Province, China[J]. Remote Sensing Letters, 2018,9(1):1-10.Abstract Bamboo forest, especially moso bamboo forest, is very important to human society. However, our ability to detect large-scale moso bamboo with optical remote sensing is still limited due to the spectral similarity with other forest species and the influence of cloud occurrence. In this study, we examined the capability of dense Landsat time series for moso bamboo forest mapping by comparing three different interpretation schemes . For each scheme, two experimental groups were further conducted to investigate the usefulness of gray-level co-occurrence matrix (GLCM) textures. Considering classification accuracy, the full-season compositing strategy was regarded as the most efficient. It was generally beneficial to include GLCM textures as input features, although their usefulness would be partially offset due to noise/correlation issues. We also investigated the roles of 15 types of auxiliary covariates in extracting moso bamboo and found some of them could enhance the classification performance significantly. With the full-season compositing scheme and crucial auxiliary covariates, an improved moso bamboo mapping performance (93.21% in overall accuracy and 73.97% in minimum accuracy) was observed within the study area. Our evaluation results are promising to provide robust guidelines for remote mapping of moso bamboo forest over large areas.

DOI

[13]
Yuan H L, Ma R H, Luo J H.Mapping orchards on plain terrains using multi-temporal medium-resolution satellite imagery[J]. Applied Engineering in Agriculture, 2015,31(3):351-362.Abstract. The ability to map the spatiotemporal distribution of orchards is essential in order to optimally utilize and manage their available resources. By analyzing multi-temporal characteristic and spectral features of medium-resolution images, the optimum features required to extract orchard data are determined and a classification rule set based on the object-based information analysis method (OBIA) is built. For those years in which images at key time points were missing, the extracted orchard data were used in the extraction process for these years as a thematic layer. The method was tested in Dangshan County, located in central China. Time series maps derived from the eight years of data over a nearly 20-year period from the Landsat Thematic Mapper (TM), Enhanced Thematic Mapper plus (ETM+) and China-HJ-1 satellite Charge Coupled Device agree well with field samples and historical records. Among the derived maps, the accuracy of orchard data derived from the 2010 HJ-1 images was estimated at about 92.9%. Compared with the common index, NDVI, the inclusion of IOPT (index for orchard in plain terrain) in the extraction process can lead to a significant improvement. When using a single-date image, the accuracies were all less than 70.0%, and when using multi-temporal images, an accuracy of 85.7% was achieved.

DOI

[14]
Li Y, Gong J, Ibrahim A N, et al.Orchard identification using landform and landscape factors based on a spatial-temporal classification framework[J]. International Journal of Remote Sensing, 2014,35(6):2118-2135.Ecological restoration measures have been undertaken in loess hilly and gully regions since the 1970s to prevent soil loss and to improve the ecological environment in those regions. Orchard construction was the main ecological measure undertaken in the Luo-Yu-Gou watershed, and in this article we propose a coupled maximum a posteriori decision rule and Markov random field (MAP-MRF) framework for orchard identification based on landform and landscape factors. Support vector machine (SVM) classification was first performed to obtain initial classification results for the years 2003 and 2008. A series of factors including landform factor, landscape factor, and the spatial emporal neighbourhood factor are used to obtain land-cover change information including the change in orchard class. Finally, field experiments were carried out in the case study region of the Luo-Yu-Gou watershed, and based on the experimental results, it was found that the quantity error and the allocation error of the classification results for 2008 were 0.0441 and 0.1037, respectively.

DOI

[15]
于新洋,张安定,侯西勇.胶东半岛果园TM影像信息的提取决策树方法[J].测绘科学,2012,37(4):57-60.

[Yu X Y, Zhang A D, Hou X Y.Decision tree classification of orchard information extraction from TM imageries in Jiaodong Peninsula of China[J]. Science of Surveying and Mapping, 2012,37(4):57-60.]

[16]
罗卫,况润元.利用环境卫星影像的东江源地区果园信息提取[J].测绘科学,2014,39(8):135-139.东江源地区是我国重要的生态 区,也是珠三角地区主要的水源地。为了快速准确地了解果园信息,文章选取东江源地区的3个县为研究区,以环境卫星的CCD影像为数据源,综合NDVI、波 段运算、地形地貌等多种辅助信息和地物的几何特征信息,构建东江源地区果园提取的决策树模型,对研究区进行果园信息的遥感提取和分类,并用混淆矩阵对分类 的精度进行评价。结果表明:东江源地区果园面积为845.7408km2,占总面积的14.08%。决策树分类模型在一定程度上提高了果园信息提取的精 度。

[Luo W, Kuang R Y.Orchard information extraction of Dongjiang source region with HJ satellite data[J]. Science of Surveying and Mapping, 2014,39(8):135-139.]

[17]
刘佳岐. 基于Landsat8遥感影像的扶风县苹果园地信息提取研究[D].西安:西北农林科技大学,2015.

[Liu J Q.Research on apple orchards information extraction of Fufeng County based on landsat8 remote sensing images[D]. Northwest A&F University, 2015.]

[18]
Reis S, Tasdemir K.Identification of hazelnut fields using spectral and Gabor textural features[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2011,66(5): 652-661.Land cover identification and monitoring agricultural resources using remote sensing imagery are of great significance for agricultural management and subsidies. Particularly, permanent crops are important in terms of economy (mainly rural development) and environmental protection. Permanent crops (including nut orchards) are extracted with very high resolution remote sensing imagery using visual interpretation or automated systems based on mainly textural features which reflect the regular plantation pattern of their orchards, since the spectral values of the nut orchards are usually close to the spectral values of other woody vegetation due to various reasons such as spectral mixing, slope, and shade. However, when the nut orchards are planted irregularly and densely at fields with high slope, textural delineation of these orchards from other woody vegetation becomes less relevant, posing a challenge for accurate automatic detection of these orchards. This study aims to overcome this challenge using a classification system based on multi-scale textural features together with spectral values. For this purpose, Black Sea region of Turkey, the region with the biggest hazelnut production in the world and the region which suffers most from this issue, is selected and two Quickbird archive images (June 2005 and September 2008) of the region are acquired. To differentiate hazel orchards from other woodlands, in addition to the pansharpened multispectral (4-band) bands of 2005 and 2008 imagery, multi-scale Gabor features are calculated from the panchromatic band of 2008 imagery at four scales and six orientations. One supervised classification method (maximum likelihood classifier, MLC) and one unsupervised method (self-organizing map, SOM) are used for classification based on spectral values, Gabor features and their combination. Both MLC and SOM achieve the highest performance (overall classification accuracies of 95% and 92%, and Kappa values of 0.93 and 0.88, respectively) when multi temporal spectral values and Gabor features are merged. High F F mathContainer Loading Mathjax values (a combined measure of producer and user accuracy) for detection of hazel orchards (0.97 for MLC and 0.94 for SOM) indicate the high quality of the classification results. When the classification is based on multi spectral values of 2008 imagery and Gabor features, similar F F mathContainer Loading Mathjax values (0.95 for MLC and 0.93 for SOM) are obtained, favoring the use of one imagery for cost/benefit efficiency. One main outcome is that despite its unsupervised nature, SOM achieves a classification performance very close to the performance of MLC, for detection of hazel orchards.

DOI

[19]
Li M, Ma L, Blaschke T, et al.A systematic comparison of different object-based classification techniques using high spatial resolution imagery in agricultural environments[J]. International Journal of Applied Earth Observations & Geoinformation, 2016,49:87-98.Geographic Object-Based Image Analysis (GEOBIA) is becoming more prevalent in remote sensing classification, especially for high-resolution imagery. Many supervised classification approaches are applied to objects rather than pixels, and several studies have been conducted to evaluate the performance of such supervised classification techniques in GEOBIA. However, these studies did not systematically investigate all relevant factors affecting the classification (segmentation scale, training set size, feature selection and mixed objects). In this study, statistical methods and visual inspection were used to compare these factors systematically in two agricultural case studies in China. The results indicate that Random Forest (RF) and Support Vector Machines (SVM) are highly suitable for GEOBIA classifications in agricultural areas and confirm the expected general tendency, namely that the overall accuracies decline with increasing segmentation scale. All other investigated methods except for RF and SVM are more prone to obtain a lower accuracy due to the broken objects at fine scales. In contrast to some previous studies, the RF classifiers yielded the best results and the k-nearest neighbor classifier were the worst results, in most cases. Likewise, the RF and Decision Tree classifiers are the most robust with or without feature selection. The results of training sample analyses indicated that the RF and adaboost. M1 possess a superior generalization capability, except when dealing with small training sample sizes. Furthermore, the classification accuracies were directly related to the homogeneity/heterogeneity of the segmented objects for all classifiers. Finally, it was suggested that RF should be considered in most cases for agricultural mapping.

DOI

[20]
Shelestov A, Lavreniuk M, Kussul N, et al.Exploring Google Earth engine platform for big data processing: classification of multi-temporal satellite imagery for crop mapping[J]. Frontiers in Earth Science, 2017,5(17):1-10.Many applied problems arising in agricultural monitoring and food security require reliable crop maps at national or global scale. Large scale crop mapping requires processing and management of large amount of heterogeneous satellite imagery acquired by various sensors that consequently leads to a ig Data problem. The main objective of this study is to explore efficiency of using the Google Earth Engine (GEE) platform when classifying multi-temporal satellite imagery with potential to apply the platform for a larger scale (e.g. country level) and multiple sensors (e.g. Landsat-8 and Sentinel-2). In particular, multiple state-of-the-art classifiers available in the GEE platform are compared to produce a high resolution (30 m) crop classification map for a large territory (~28,100 km2 and 1.0 M ha of cropland). Though this study does not involve large volumes of data, it does address efficiency of the GEE platform to effectively execute complex workflows of satellite data processing required with large scale applications such as crop mapping. The study discusses strengths and weaknesses of classifiers, assesses accuracies that can be achieved with different classifiers for the Ukrainian landscape, and compares them to the benchmark classifier using a neural network approach that was developed in our previous studies. The study is carried out for the Joint Experiment of Crop Assessment and Monitoring (JECAM) test site in Ukraine covering the Kyiv region (North of Ukraine) in 2013. We found that Google Earth Engine (GEE) provides very good performance in terms of enabling access to the remote sensing products through the cloud platform and providing pre-processing; however, in terms of classification accuracy, the neural network based approach outperformed support vector machine (SVM), decision tree and random forest classifiers available in GEE.

DOI

[21]
王九中,田海峰,邬明权,等.河南省冬小麦快速遥感制图[J].地球信息科学学报,2017,19(6):846-853.

[Wang J Z, Tian H F, Wu M Q, et al.Rapid mapping of winter wheat in Henan Province[J]. Journal of Geo-information Science, 2017,19(6):846-853.]

[22]
Farr T G, Rosen P A, Caro E, et al.The shuttle radar topography mission[J]. Reviews of Geophysics, 2007,45(RG2004).The Shuttle Radar Topography Mission produced the most complete, highest-resolution digital elevation model of the Earth. The project was a joint endeavor of NASA, the National Geospatial-Intelligence Agency, and the German and Italian Space Agencies and flew in February 2000. It used dual radar antennas to acquire interferometric radar data, processed to digital topographic data at 1 arc sec resolution. Details of the development, flight operations, data processing, and products are provided for users of this revolutionary data set.

DOI

[23]
Li X, Gong P, Liang L.A 30-year (1984-2013) record of annual urban dynamics of Beijing City derived from Landsat data[J]. Remote Sensing of Environment, 2015,166:78-90.61An annual sequence of urban land has been produced in Beijing over a 30-year period.61Many Landsat images have been employed to make full use of their temporal contexts.61A temporal consistency check was conducted to make the sequence more reasonable.61The growth rates are different in Beijing during the past three decades.

DOI

[24]
Tucker C J.Red and photographic infrared linear combinations for monitoring vegetation[J]. Remote Sensing of Environment, 1979,8(2):127-150.In situ collected spectrometer data were used to evaluate and quantify the relationships between various linear combinations of red and photographic infrared radiances and experimental plot biomass, leaf water content, and chlorophyll content. The radiance variables evaluated included the red and photographic infrared (IR) radiance and the linear combinations of the IR/red ratio, the square root of the IR/red ratio, the IR-red difference, the vegetation index, and the transformed vegetation index. In addition, the corresponding green and red linear combinations were evaluated for comparative purposes. Three data sets were used from June, September, and October sampling periods. Regression analysis showed the increased utility of the IR and red linear combinations vis- -vis the same green and red linear combinations. The red and IR linear combinations had 7% and 14% greater regression significance than the green and red linear combinations for the June and September sampling periods, respectively. The vegetation index, transformed vegetation index, and square root of the IR/red ratio were the most significant, followed closely by the IR/red ratio. Less than a 6% difference separated the highest and lowest of these four ER and red linear combinations. The use of these linear combinations was shown to be sensitive primarily to the green leaf area or green leaf biomass. As such, these linear combinations of the red and photographic IR radiances can be employed to monitor the photosynthetically active biomass of plant canopies.

DOI

[25]
Wilson E H, Sader S A.Detection of forest harvest type using multiple dates of Landsat TM imagery[J]. Remote Sensing of Environment, 2002,80(3):385-396.A simple and relatively accurate technique for classifying time-series Landsat Thematic Mapper (TM) imagery to detect levels of forest harvest is the topic of this research. The accuracy of multidate classification of the normalized difference vegetation index (NDVI) and the normalized difference moisture index (NDMI) were compared and the effect of the number of years (1–3, 3–4, 5–6 years) between image acquisition on forest change accuracy was evaluated. When Landsat image acquisitions were only 1–3 years apart, forest clearcuts were detected with producer's accuracy ranging from 79% to 96% using the RGB-NDMI classification method. Partial harvests were detected with lower producer's accuracy (55–80%) accuracy. The accuracy of both clearcut and partial harvests decreased as time between image acquisition increased. In all classification trials, the RGB-NDMI method produced significantly higher accuracies, compared to the RGB-NDVI. These results are interesting because the less common NDMI (using the reflected middle infrared band) outperformed the more popular NDVI. In northern Maine, industrial forest landowners have shifted from clearcutting to partial harvest systems in recent years. The RGB-NDMI change detection classification applied to Landsat TM imagery collected every 2–3 years appears to be a promising technique for monitoring forest harvesting and other disturbances that do not remove the entire overstory canopy.

DOI

[26]
Huete A R.A soil-adjusted vegetation index (SAVI)[J]. Remote Sensing of Environment, 1988,25(3):295-309.A transformation technique is presented to minimize soil brightness influences from spectral vegetation indices involving red and near-infrared (NIR) wavelengths. Graphically, the transformation involves a shifting of the origin of reflectance spectra plotted in NIR-red wavelength space to account for first-order soil-vegetation interactions and differential red and NIR flux extinction through vegetated canopies. For cotton ( Gossypium hirsutum L. var DPI-70) and range grass ( Eragrostics lehmanniana Nees) canopies, underlain with different soil backgrounds, the transformation nearly eliminated soil-induced variations in vegetation indices. A physical basis for the soil-adjusted vegetation index (SAVI) is subsequently presented. The SAVI was found to be an important step toward the establishment of simple lobal that can describe dynamic soil-vegetation systems from remotely sensed data.

DOI

[27]
Haralick R M. Texture features for image classification[J]. Systems Man & Cybernetics IEEE Transactions on, 1973,smc-3(6):610-621.CiteSeerX - Scientific documents that cite the following paper: Textural features for image classification

DOI

[28]
Song C, Woodcock C E, Seto K C, et al.Classification and change detection using Landsat TM data: When and how to correct atmospheric effects?[J]. Remote Sensing of Environment, 2001,75(2):230-244.

DOI

[29]
Breiman L.Random Forests[J]. Machine Learning, 2001,45(1):5-32.

DOI

[30]
Belgiu M, Dr Gu L.Random forest in remote sensing: A review of applications and future directions[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016,114(Supplement C): 24-31.A random forest (RF) classifier is an ensemble classifier that produces multiple decision trees, using a randomly selected subset of training samples and variables. This classifier has become popular within the remote sensing community due to the accuracy of its classifications. The overall objective of this work was to review the utilization of RF classifier in remote sensing. This review has revealed that RF classifier can successfully handle high data dimensionality and multicolinearity, being both fast and insensitive to overfitting. It is, however, sensitive to the sampling design. The variable importance (VI) measurement provided by the RF classifier has been extensively exploited in different scenarios, for example to reduce the number of dimensions of hyperspectral data, to identify the most relevant multisource remote sensing and geographic data, and to select the most suitable season to classify particular target classes. Further investigations are required into less commonly exploited uses of this classifier, such as for sample proximity analysis to detect and remove outliers in the training samples.

DOI

[31]
Gong P, Wang J, Yu L, et al.Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data[J]. International Journal of Remote Sensing, 2013,34(7):2607-2654.We have produced the first 30 m resolution global land-cover maps using Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) data. We have classified over 6600 scenes of Landsat TM data after 2006, and over 2300 scenes of Landsat TM and ETM+ data before 2006, all selected from the green season. These images cover most of the world's land surface except Antarctica and Greenland. Most of these images came from the United States Geological Survey in level L1T (orthorectified). Four classifiers that were freely available were employed, including the conventional maximum likelihood classifier (MLC), J4.8 decision tree classifier, Random Forest (RF) classifier and support vector machine (SVM) classifier. A total of 91,433 training samples were collected by traversing each scene and finding the most representative and homogeneous samples. A total of 38,664 test samples were collected at preset, fixed locations based on a globally systematic unaligned sampling strategy. Two software tools, Global Analyst and Global Mapper developed by extending the functionality of Google Earth, were used in developing the training and test sample databases by referencing the Moderate Resolution Imaging Spectroradiometer enhanced vegetation index (MODIS EVI) time series for 2010 and high resolution images from Google Earth. A unique land-cover classification system was developed that can be crosswalked to the existing United Nations Food and Agriculture Organization (FAO) land-cover classification system as well as the International Geosphere-Biosphere Programme (IGBP) system. Using the four classification algorithms, we obtained the initial set of global land-cover maps. The SVM produced the highest overall classification accuracy (OCA) of 64.9% assessed with our test samples, with RF (59.8%), J4.8 (57.9%), and MLC (53.9%) ranked from the second to the fourth. We also estimated the OCAs using a subset of our test samples (8629) each of which represented a homogeneous area greater than 500 m 500 m. Using this subset, we found the OCA for the SVM to be 71.5%. As a consistent source for estimating the coverage of global land-cover types in the world, estimation from the test samples shows that only 6.90% of the world is planted for agricultural production. The total area of cropland is 11.51% if unplanted croplands are included. The forests, grasslands, and shrublands cover 28.35%, 13.37%, and 11.49% of the world, respectively. The impervious surface covers only 0.66% of the world. Inland waterbodies, barren lands, and snow and ice cover 3.56%, 16.51%, and 12.81% of the world, respectively.

DOI

[32]
Zhu Z.Change detection using landsat time series: A review of frequencies, preprocessing, algorithms, and applications[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2017,130:370-384.Abstract The free and open access to all archived Landsat images in 2008 has completely changed the way of using Landsat data. Many novel change detection algorithms based on Landsat time series have been developed We present a comprehensive review of four important aspects of change detection studies based on Landsat time series, including frequencies, preprocessing, algorithms, and applications. We observed the trend that the more recent the study, the higher the frequency of Landsat time series used. We reviewed a series of image preprocessing steps, including atmospheric correction, cloud and cloud shadow detection, and composite/fusion/metrics techniques. We divided all change detection algorithms into six categories, including thresholding, differencing, segmentation, trajectory classification, statistical boundary, and regression. Within each category, six major characteristics of different algorithms, such as frequency, change index, univariate/multivariate, online/offline, abrupt/gradual change, and sub-pixel/pixel/spatial were analyzed. Moreover, some of the widely-used change detection algorithms were also discussed. Finally, we reviewed different change detection applications by dividing these applications into two categories, change target and change agent detection.

DOI

[33]
李自茂,钟八莲,孙剑斌.2013赣南脐橙产业发展报告[M].北京:经济管理出版社,2014.

[Li Z M, Zhong B L, Sun J B.Report on the development of Gannan navel orange industry (2013)[M]. Beijing: Economic Management Press, 2014.]

Options
Outlines

/

〈 〉