EN中文

Journal of Geo-information Science >

2016 , Vol. 18 >Issue 5: 708 - 717

DOI: https://doi.org/10.3724/SP.J.1047.2016.00708

Orginal Article

Study on the Crop Classification and Planting Area Estimation at Land Parcel Scale Using Multi-sources Satellite Data

  • HUANG Qiting , 1, 2, * ,
  • QIN Zelin 3 ,
  • ZENG Zhikang 3
Expand
  • 1. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China
  • 2. Graduate University of Chinese Academy of Sciences, Beijing 100049, China
  • 3. Scientific and Technological Information Research Institute, Guangxi Academy of Agricultural Sciences, Nanning, 530007, China;
*Corresponding author: HUANG Qiting, E-mail:

Received date: 2015-12-16

  Request revised date: 2016-03-31

  Online published: 2016-05-10

Copyright

《地球信息科学学报》编辑部 所有
Fold

Abstract

To reduce the missing of remotely sensed data in the spatio-temporal coverage of the cloudy/rainy region and to further meet the urgent need for crop planting information at farmland parcel scale, a method of crop type identification and planting area estimation at parcel scale was developed in this paper by synergistically utilizing the multi-sources satellite imagery,with the support of remote sensing Tupu recognization theory. This method consists of three steps: firstly, based on the high resolution imagery, the objects of farmland parcel with exact boundary were extracted. Secondly, with the effective-data processing technology and the spectral indices calculation based on the multi-temporal medium resolution imagery, the fragmentary effective data was acquired and the time-series data for each object was further obtained. Finally, by constructing a multi-dimensional feature space with the help of time series analysis incorporating the crops’phenological feature, the crop types and their corresponding planting areas were mapped using the Decision Tree classifier. This method had been tested in Ningyuan county, Hunan Province, China. The results showed that, this method can precisely map the different rice types and corresponding planting areas at the farmland parcel scale. The user accuracy of the three rice types, i.e., the early double-season, single-season and late double-season rice,was 94.33%, 90.76 and 95.95%, respectively, and the overall accuracy was 92.51% with a Kappa coefficient of 0.90. The derived area accuracy of these three rice types also reached 93.37%, 91.23% and 95.42%,respectively. This experiment illustrated the effectiveness and usefulness of the proposed method and also provided a salutary lesson for the finely planting information extraction of other crops.

Key words： remote sensing; Tupu recognization theory; crop identification; land parcel scale

Cite this article

HUANG Qiting , QIN Zelin , ZENG Zhikang . Study on the Crop Classification and Planting Area Estimation at Land Parcel Scale Using Multi-sources Satellite Data[J]. Journal of Geo-information Science, 2016 , 18(5) : 708 -717 . DOI: 10.3724/SP.J.1047.2016.00708

1 引言

遥感技术以其宏观、快速和准确等特点,已成为农业资源调查的有力手段,且利用遥感技术获取农作物种植面积信息的应用研究日益广泛[1-4]。目前作物面积提取的遥感数据源主要是AVHRR、MODIS等低分辨率数据[5-7]和TM/ETM、HJ等中分辨率数据[8-9]。低分辨率数据主要利用其高重访频率的特点,以作物物候或生长关键期的光谱特征以及环境变化特点（如水稻移栽期田块被浅水浸没等）为依据,采用基于植被指数时间序列的分析方法进行分类识别及面积计算[10-12]。相对而言,中分辨率数据主要依靠较高空间分辨率的优势,利用目标作物与其他地类在影像上的光谱特征差异进行分类提取[13-14]。总体而言,低分辨率数据由于空间尺度较粗,混合像元现象突出,导致其解译精度有限;中分辨率数据虽然具有相对较高的空间分辨率,但其重访周期长且受影像质量的严重制约,单一传感器影像难以满足多时相无云观测的要求,晴空影像的可获取程度决定了时相的选取与方法的有效性,因此目前大多利用1~2个时相来进行作物识别及其面积估算。另外,无论是中分辨率还是低分辨率的作物识别与面积估算,大多在像元尺度上进行,一般先经过影像分类识别作物,然后通过面积估算模型进行面积求算以处理混合像元中复杂的地物组分构成,难以达到种植面积的精确测量。随着精准农业管理、农业补贴以及农业保险核查等领域业务的发展,对地块尺度的作物种植面积信息提出了越来越迫切的需求,需要从遥感数据获取及利用方式上进行新的探索尝试。
近年来,中国已成功发射HJ1-A/B、GF1-WFV、CBERS-04等多颗中分辨率卫星,同时ZY-3、ZY-02C、GF2等米级高空间分辨率数据也不断涌现,为实现多尺度、高时空分辨率覆盖的对地观测提供了可能。如何挖掘数据潜力,最大程度发挥不同尺度数据源的优势,是当前遥感应用中亟需解决的问题,其实质是在海量多源遥感数据条件下如何实现特定专题信息的准确获取。遥感信息“图-谱”认知理论为多源数据条件下遥感信息的准确提取提供了理论指导框架,认为遥感图像具有“图”和“谱”的双重特性：“图”是指遥感影像中地物呈现的精细几何图式信息（如结构、形状、格局等）;“谱”则是地物对象蕴含的光谱、属性和规律等多层次的特征和知识,通过“图谱合一”的协同耦合,从多角度“立体”地刻画地物全貌特征,最终达到对地物的全面和准确认知,进而实现遥感地物信息的精准提取[15]。
本文在遥感信息“图-谱”认知理论支持下,提出了基于多星数据协同的地块尺度作物识别与面积估算方法,在有效像元处理技术基础上,通过高分辨率数据与多源中分辨率时序数据的协同,将从高分辨率影像提取的农田地块精细“图”信息与从多源中分辨率时序影像获得的作物光谱变化、物候特征等“谱”信息进行有机融合,构建反映地块作物生长过程的时间序列与作物识别模型,实现农田地块尺度的作物识别与面积计算。最后,以湖南省宁远县的不同水稻类型种植面积提取为例,探讨有效像元处理技术在南方典型耕地和多云气候条件下构建农作物生长过程时序的能力,并在此基础上探索了地块尺度作物生长过程特征及作物识别方法,最后对本文方法在作物识别和面积估算方面的精度水平与应用潜力进行了分析评价。

2 研究方法

在统一时空框架下,本文以高分辨率影像作为“图信息”来源提取精细的农田地块边界,并在地块控制下融入多源中分辨率时序影像的光谱及时序物候特征等“谱信息”,进而以地块为单位构建作物判别模型,实现作物的识别与面积求算。此外,为提高中分辨率数据的时空覆盖度,采用有效像元处理技术对中分辨率时序影像进行有效化处理,充 分挖掘影像的可用信息,使之服务于遥感信息提取。提取过程主要由4部分构成：（1)精细地块边界获取;（2）中分辨率数据有效化处理;（3）地块特征提取;（4）特征分析与分类。具体技术流程如图1 所示。
Fig.1 Flow chart of crop planting area extraction based on multi-source satellite data

图1 多源遥感数据协同的作物种植面积提取流程图

2.1 精细地块边界获取

资源三号等米级分辨率影像具有适中的观测尺度,在该尺度上地块边界清晰,同时地块内部的作物细节得到适当综合,表现出良好的光谱均质性,为地块的准确、完整提取提供有利条件。地块边界的提取采用影像分割与手工编辑勾绘相结合的方式,在道路网和耕地范围矢量辅助下,对耕地区域进行多尺度分割,而后对分割边界进行基于目视的手工编辑勾绘与整体平滑处理,最终获得农田地块的完整边界矢量,如图2所示。尽管地块边界的首次提取需较多人工参与,但因地块结构相对稳定,提取结果具有较高复用性,有利于地块作物信息的快速更新。
Fig.2 The extraction result of farmland parcel

图2 农田地块边界提取结果

2.2 中分辨率数据有效化处理

数据有效化处理是指将常规遥感图像处理中视为无用数据（云量>30%）的影像也利用起来,在几何和辐射校正基础上,提取云影间的有效像元区域,充分发掘影像的可用信息,通过“碎片化”的数据利用方式达到提高遥感观测时空覆盖度的目的。其中,云影的有效检测是数据有效化处理的关键。本文基于无云ZY3影像底图,通过RGB到YCbCr的空间变换对中分辨率影像的云影区域进行融合增强,然后采用Otsu方法自动寻找最优阈值实现云影区域的有效检测（具体方法详见文献[16]）。云影检测结果经检查、修正后与对应影像进行掩膜处理,生成影像有效数据,并进一步根据式（1）计算、获取碎片化的NDVI数据集。数据有效化过程如图3所示。
NDVI = ρ NIR - ρ RED ρ NIR + ρ RED （1）
式中： ρ NIR 为近红外波段反射率; ρ RED 为红波段反射率。
Fig.3 Demonstration of the processed effective imagery data

图3 影像数据有效化过程示意图

注：图（c）-（d）中黑色部分为无数据区域

2.3 地块特征提取

（1）多传感器植被指数归一化
为减轻不同卫星传感器间由于波段设置、观测几何等差异导致的植被指数的偏差,本文采用线性回归方法,以GF-WVF1为基准,对其它传感器的植被指数求取回归模型进行拟合换算,实现不同NDVI数据间的相对归一化。不同传感器的植被指数间线性拟合归一化的原理方法可参见文献[17]、[18]。首先选择目标传感器与待纠正传感器在同区域、时相相近的晴空影像对,从重叠区中选取约100个纯净样点构建二者的回归模型,然后基于该模型对待校正数据进行拟合换算。样点大小视像元相对大小而定：GF-WVF之间以单个像元值为抽样值;GF-WVF1与HJ-1A/B之间,样点为半径16 m的圆,GF-WVF1以圆内4个像元均值为抽样值,而HJ-1A/B则以单个像元为样值;GF-WVF1和Landsat8-OLI传感器间的校正与HJ-1A/B的校正类似。由不同传感器与GF-WVF1植被指数的线性回归模型（表1）可知,各传感器拟合方程的相关系数和均方根误差均较为理想,基本满足分析应用需求。
Tab.1 Regression models of GF-WVF1 and other satellite sensors

表1 GF-WVF1与其它传感器的NDVI回归模型

传感器 拟合方程 R2 RMSE
GF1-WFV2 Y=0.7690X+0.1699 0.8977 0.0195
GF1-WFV3 Y=0.8233X+0.1285 0.6362 0.0416
GF1-WFV4 Y=0.9195X+0.0446 0.7975 0.0339
HJ1A-CCD1 Y=1.0673X+0.0256 0.9059 0.0209
HJ1A-CCD2 Y=1.0858X+0.0512 0.9018 0.0211
HJ1B-CCD1 Y=0.9157X+0.1765 0.7275 0.0194
HJ1B-CCD2 Y=0.8469X+0.1928 0.8064 0.0328
Landsat8-OLI Y=0.9018X-0.0387 0.7498 0.0683

注：Y为GF-WVF1传感器的NDVI值;X为其它传感器NDVI值

（2）时间序列获取与重建
基于归一化的NDVI有效数据集,以地块NDVI均值为取值,获取地块初始时间序列。由于NDVI有效数据集为无云影像数据块,有效地解决了云污染引起的时序异常值问题,然而其非等时间间隔的特点也导致了不同地块的初始序列在观测频度分布上有所差异。为了保证特征提取结果的准确性和一致性,利用SPLINE函数进行拟合插值,生成如图4（a）所示间隔为10 d的时间序列,并进一步采用Savitzky-Golay滤波（又称S-G滤波）对时间序列进行重建。S-G滤波的窗口大小和多项式次数分别为4和3,时间序列重建效果如图4（b）所示。
Fig.4 The contrast of NDVI time-series reconstructions for crops

图4 作物NDVI时间序列重建效果对比

（3）地块光谱及物候特征提取
作物在不同生育阶段表现出不同的光谱特征,在时序上表现为NDVI随时间的变化曲线,曲线的上升和下降与作物生长和成熟等过程相对应,可以有效地反映作物的生育和物候特点。由于物候特征和结构长势的差异,不同作物的NDVI曲线具有不同形态和量值,可根据曲线的变化特点获取遥感物候参数对作物进行识别。
目前,已有不少学者基于中低分辨率时序影像开展了作物物候监测及作物种植信息的提取研究[19-20]。本文在前人研究基础上,采用一系列光谱统计和物候特征,构建地块多维特征空间用以支持作物识别分类。光谱统计特征包括NDVI最大值、NDVI最小值、NDVI月均值和年均值;同时基于动态阈值法[21]进行了①曲线峰数、②生长季开始时刻、③生长季结束时刻、④生长季长度及⑤凋零速率等遥感物候参数的提取(图5)。由上述光谱和物候特征构成的特征空间,丰富了作物属性表达的特征维度,有助于提高作物遥感识别准确度。
Fig. 5 Remotely sensed phonological features derived from NDVI curve

图5 基于NDVI曲线的遥感物候特征示意图

2.4 特征分析与分类

面对地块特征空间中数量众多的光谱和物候特征,选取合适的变量组合与分类方法尤为关键。有研究表明[22],决策树方法具有良好的灵活性和鲁棒性,不仅可以处理光谱、空间和高程等多源数据,还可以有效地处理大量高维数据和非线性关系。本文在对研究区主要作物生长特性和物候特征进行分析的基础上,选取对作物识别起主要贡献的特征组合,通过样本统计的方式获取各变量的阈值,构建基于规则的决策树模型,进而对目标作物进行分类识别。

3 实验与结果分析

3.1 研究区与数据处理

本文以湖南省宁远县为研究区,对早、中、晚稻等不同水稻类型进行分类和种植面积提取。宁远县位于湖南南部,地处110°42′~112°27′E,25°11′~26°08′N,属亚热带季风湿润区,境域四面环山,受气候和地形影响,常年多云雨天气。研究区耕地为中等破碎度,面积超过300 m2的地块数占比>90%,地块平均大小介于860~1450 m2之间;耕作制度以早稻-晚稻、中稻-油菜轮作为主,水稻、油菜、烟草和蔬菜是主要的农作物类型,主要作物的套（间）作比例小。
本研究所用数据包括遥感数据与辅助矢量数据2类。遥感数据包括ZY-3米级分辨率影像和由GF1-WFV、HJ-1A/B构成的多传感器中分时序影像数据集,具体信息如表2所示。ZY-3全色/多光谱影像拼接后可在研究区形成无云覆盖,其成像时间虽然与中分辨率影像不同,但由于农田地块本身具有相对稳定性,在1-2年内边界变化基本可忽略不计,因此在地块边界提取方面仍不失其有效性;中分辨率时序数据为2014年3-10月的多传感器数据共33景,可实现每月至少2次的多期观测。辅助数据为研究区2 m分辨率道路网和耕地范围矢量。
Tab. 2 Imagery information of this study

表2 研究采用的影像信息

时间 中心经纬度 传感器 时间 中心经纬度 传感器
2014-03-17 E113.1_N25.6 GF1_WFV3 2014-09-21 E113.2_N26.3 GF1_WFV1
2014-03-26 E111.0_N24.6 GF1_WFV4 2014-09-25 E111.8_N24.7 GF1_WFV1
2014-03-26 E110.5_N25.2 GF1_WFV4 2014-09-26 E110.9_N25.9 GF1_WFV1
2014-04-04 E112.8_N27.1 HJ1B-CCD2 2014-10-04 E112.8_N25.6 GF1_WFV3
2014-04-04 E111.0_N26.8 HJ1B-CCD2 2014-10-08 E111.6_N25.9 GF1_WFV2
2014-04-14 E110.9_N27.4 HJ1A-CCD2 2014-10-16 E110.8_N26.3 GF1_WFV1
2014-05-01 E111.3_N25.9 GF1_WFV2 2014-10-16 E112.8_N25.9 GF1_WFV2
2014-05-01 E113.1_N25.6 GF1_WFV3 2014-10-24 E111.6_N24.6 GF1_WFV1
2014-06-13 E110.6_N25.2 HJ1B-CCD2 2014-10-24 E111.9_N26.3 GF1_WFV1
2014-06-13 E110.8_N26.3 HJ1A-CCD1 2014-11-14 E111.1_N25.9 GF1_WFV2
2014-06-15 E112.0_N25.9 GF1_WFV2 2014-11-18 E110.8_N26.3 HJ1B-CCD1
2014-07-10 E112.1_N25.6 GF1_WFV3 2014-11-22 E112.2_N25.9 GF1_WFV2
2014-07-18 E111.6_N25.9 GF1_WFV2 2012-10-01 E111.8 _N25.5 ZY3_NAD
2014-07-30 E111.1_N24.6 GF1_WFV1 2012-10-01 E111.8 _N25.5 ZY3_MUX
2014-07-30 E111.5_N26.3 GF1_WFV1 2012-10-01 E111.9 _N25.9 ZY3_NAD
2014-08-03 E111.7_N24.6 GF1_WFV1 2012-10-01 E111.9 _N25.9 ZY3_MUX
2014-08-03 E112.1_N26.3 GF1_WFV1 2013-08-02 E112.2_N 25.1 ZY3_NAD
2014-08-29 E109.9_N26.8 HJ1A-CCD1 2013-08-02 E112.2_N 25.1 ZY3_MUX
2014-09-01 E112.2_N25.9 GF1_WFV2 2013-08-02 E112.3_N 25.5 ZY3_NAD
2014-09-04 E112.0_N26.4 HJ1B-CCD1 2013-08-02 E112.3_N 25.5 ZY3_MUX
2014-09-21 E112.8_N24.6 GF1_WFV1
数据处理主要包括辐射校正、几何校正以及地块相关特征获取。辐射校正首先将影像DN值定标为辐亮度,然后采用6S辐射传输模型进行大气校正,实现DN值到地表反射率的转换;几何校正在对ZY3影像进行正射和融合处理基础上,以2 m ZY3融合影像为参考对多源中分辨率数据进行几何纠正,纠正误差控制在1个GF-WFV像元（16 m）内。最后,基于ZY-3融合影像和中分辨率数据集进行精细地块边界的提取、中分辨率数据有效化以及地块光谱和时序特征计算,获得作物识别所需的多维特征空间。

3.2 主要作物生长过程及物候分析

图6为经平滑处理的作物生长曲线,由图可见,作物生育周期一般历经从生长较缓慢的苗期到生长速率急剧增大的旺长期,再到鼎盛期,而后逐渐衰退至成熟期的若干阶段。同时,不同作物曲线在峰/谷分布、峰宽等方面差异明显,反映了作物在发育鼎盛期、成熟期和生长季长度等物候特征上的差别。结合表3作物物候节律可知,研究区油菜3-4月处于高覆被的生长盛期,至5月成熟收割,此时其它作物未播种或处于苗期低覆盖状态,据此可识别油菜;双季稻（早稻-晚稻）分别在6月下旬和10上旬达到鼎盛期峰值,与7-8月间因收割/移栽形成的波谷构成特征性的双峰形态,易于判别;中稻6月中下旬完成移栽,地表覆被由低转高,与早稻、烟叶等其它处于生长盛期作物表现相反变化趋势;烟叶从3月上旬移栽至8月下旬采摘完成,生育期长达6个月,通过移栽和收割时点及生长时长的判定可对烟叶进行有效识别。以上分析表明,研究区作物在时间维上具有较强可分性,可通过遥感物候参数及NDVI月均值等特征的适当组合来刻画作物生长过程的变化特点,实现不同作物的有效识别。
Fig.6 NDVI curves of different crops

图6 不同作物的NDVI生长曲线

Tab.3 Phenological calendar of the main crops in the study areas

表3 研究区主要作物物候历

作物 月份
3 4 5 6 7 8 9 10
早稻 移栽期 分蘖期 幼穗发育期 成熟期
晚稻 移栽期 分蘖期 幼穗发育期 成熟期
中稻 移栽期 分蘖期 幼穗发育期 成熟期
烟叶 返苗期 伸根期 旺长期 成熟期
油菜 开花期 开花期 成熟期 苗期

3.3 水稻识别分类

为了确定实验区作物的真实分布特征,于2014年7月中旬和9月中旬开展了野外实地调查（图7）,采用GPS样区定位与地块详查方式获取了水稻地块样本共1407个,烟草、玉米和抛荒地等其他类样本461个。将耕地植被分为早稻、中稻、晚稻和其他地类共4类,并根据3.2节的作物生长过程分析,选取NDVI月均值、曲线峰数、凋零速率、生长季开始/结束时点及生长季长度共6个特征变量构建决策树分类规则。进一步将地块样本分为训练样本和验证样本2部分,针对不同类别基于训练样本计算、统计各特征变量的均值和取值范围,通过变量值的调整使各类别具有较好的可区分性,人工获取各特征变量的分类阈值,实现不同类型水稻的分类,并基于验证样本对分类结果进行精度评价。最终分别建立了早稻提取模型（式（2））、中稻提取模型（式（3））、晚稻提取模型（式（4））。
Fig.7 The location of the experimental area

图7 研究区位置图

NDV I mean 6 ≥ 0.6 Nu m Peak ≥ 2 ΔNDV I Slope ≥ 0.86 100 ≤ GrowBeg i nDate ≤ 120 190 ≤ GrowEndDate ≤ 210 GrowPeriod ≤ 100 Days （2）
NDV I mean 8 ≥ 0.6 Nu m Peak ≥ 1 ΔNDV I Slope ≥ 0.86 160 ≤ GrowBeginDate ≤ 180 250 ≤ GrowEndDate ≤ 270 GrowPeriod ≤ 100 Days （3）
NDV I mean 9 ≥ 0.6 Nu m Peak ≥ 2 ΔNDV I Slope ≥ 0.86 190 ≤ GrowBeginDate ≤ 210 280 ≤ GrowEndDate ≤ 300 GrowPeriod ≤ 100 Days （4）
式中： NDV I mean 6 、 NDV I mean 8 、 NDV I mean 9 分别为6、8、9月的NDVI月均值; Nu m Peak 为时序曲线峰数,与作物熟制或轮作制度相关,每茬作物对应一个波峰; ΔNDV I Slope 为凋零速率,其值大于0.86时表明地表覆被发生了剧烈变化,可区分自然植被和农作物; GrowBeginDate 和 GrowEndDate 分别为生长季开始和结束时间,单位为一年当中的第几天; GrowPeriod 为以天数为单位的生长季长度。

3.4 结果与评价

分类获得的各水稻类型分布如图8（a）、（b）所示,研究区水稻分布具有一定规律性,中稻主要集中于中、西部平坦地区,早、晚双季稻则主要分布于中部偏南北方向,而山势险峻的南北两端鲜有水稻种植。从图8（c）的局部放大图可看出,本文方法可在农田地块尺度上对水稻种植分布的空间和属性进行精细表达,不仅有利于后期验证,同时还有利于与权属人、药肥施用量等其它信息作叠加分析,为更深层次的精准化应用提供数据基础。以下从分类精度和面积精度2方面对方法的有效性进行评价分析。
Fig. 8 The distribution of rice types

图8 各水稻类型分布图

（1）分类精度评价
分类误差矩阵如表4所示,本文方法的总体分类精度达到92.51%,Kappa系数为0.90,整体分类效果较为理想。早、晚稻的分类精度均为94%以上,中稻的用户和制图精度接近于90%,不同类型水稻均得到较有效的识别。
Tab.4 Confusion matrix for rice classification

表4 水稻分类混淆矩阵

类别 早稻 中稻 晚稻 其他地类 样本总数 用户精度/（%）
早稻 416 10 0 15 441 94.33
中稻 3 285 7 19 314 90.76
晚稻 0 10 379 6 395 95.95
其他地类 18 12 7 241 278 86.69
样本总数 437 317 393 281 1428
制图精度/（%） 95.19 89.91 96.44 85.77
总体精度/（%） 92.51
Kappa系数 0.90
（2）面积精度评价
在分类结果的基础上,对同一类型的水稻地块进行面积累加统计,获得研究区各类型水稻面积。如表5所示,早、中、晚稻的遥感提取面积分别为12.96、11.13和13.32千hm2,以同年国家统计局湖南调查总队公布的各类型水稻面积为面积精度评价标准,早、中、晚稻的面积精度分别为93.37%、91.23%和95.42%,平均精度达到93.43%,均取得了较高的面积提取精度。
Tab.5 Accuracy of rice planting area extraction

表5 水稻种植面积提取精度

类别 遥感提取面积/千hm2 统计数据/千hm2 面积精度/（%）
早稻 12.96 13.88 93.37
中稻 11.13 12.20 91.23
晚稻 13.32 13.96 95.42
总计 37.41 — —
平均 — — 93.43

4 结论与讨论

本文提出了一种图谱认知框架下基于高时空分辨率多星数据协同的精细化作物识别与种植面积提取方法,并以湖南省宁远县为例,开展了地块尺度下不同类型水稻的分类识别与面积提取研究,其总体分类精度和平均面积提取精度分别达到92.51%和93.43%,Kappa系数为0.9,表明了方法的有效性。研究还得出以下结论：
（1）以地块作为分类与面积计算的基本单元,避免了象元级分类中的“椒盐”现象,提高了分类和面积测算精度及结果的可验证性;同时,精准的地块边界比传统面向对象的不规则图斑更具有自然和社会属性意义,有望将遥感信息的分析和应用尺度拓展到地块级别;
（2）多星数据协同及“碎片化”有效数据利用方式显著提高了遥感数据的时空覆盖度,可为多云雨气候条件下的作物种植信息提取和长势监测提供高分辨率、高频度的对地观测信息支持。
（3）通过高时空分辨率数据的“图-谱”信息协同,在构建地块时间序列基础上引入作物物候信息,可有效增加作物的类别可分性,提高作物识别精度。
然而,地块破碎度和作物种植模式对分类精度也会带来影响,目前本文方法适用于地块破碎和套（间）作程度不高的地区,在小地块和作物混杂严重区域应考虑引入像元解混的处理方法。总体而言,本文仍存在以下有待改进之处：（1）影像分割方法优化,使地块对象与农田自然边界更加吻合,以提高地块获取精度和效率;（2）零碎地块及多作物混杂地块仍难以处理,需加强尺度转换与像元分解模型的研究,以有效解决由此带来的混合像元问题;（3）本研究主要以物候节律信息为主,未来工作将考虑引入光谱、地形等多元特征,研究自适应的特征选取与阈值优化方法,构建多特征联合的自动化作物识别模型,进一步提高作物种植信息提取的精度和智能化水平。

The authors have declared that no competing interests exist.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Xiao X, Boles S, Liu J, et al.Mapping paddy rice agriculture in southern China using multi-temporal MODIS images[J]. Remote Sensing of Environment, 2005,95(4):480-492.Information on spatial extent and seasonality of inundation and paddy rice fields are needed for water resource management, trace gases emission, and food security. In this study we reported an effort to use images from Moderate Resolution Imaging Spectroradiometer (MODIS) sensor onboard NASA EOS Terra satellite to map inundation and paddy rice fields in southern China. Paddy rice fields are characterized by a period of inundation and open canopy (a mixture of surface water and rice crops). We developed a temporal profile analysis procedure that uses time series data of improved vegetation indices to identify and map inundation and paddy rice fields. The MODIS-based algorithm uses both Enhanced Vegetation Index (EVI) and Land Surface Water Index (LSWI), and excludes those pixels that are covered by cloud and snow from the analysis. Permanent water body mask and digital elevation model were also used in the analysis. Using multi-temporal 8-day composite of MODIS images at 500-m spatial resolution in 2002, we generated a map of inundation and paddy rice fields in southern China. The MODIS-derived paddy rice map was compared with the other datasets of paddy rice agriculture, including the paddy rice map derived from analysis of Landsat ETM+ images in 1999/2000. The results from the comparison have indicated that the MODIS-based algorithm could potentially be applied at large spatial scale for mapping and monitoring of inundation and paddy rice agriculture.

DOI

[2]
Xiao X M, Boles S, Frolking S, et al.Mapping paddy rice agriculture in south and southeast Asia using multi-temporal MODIS images[J]. Remote Sensing of Environment, 2006,100(1):95-113.In this paper, we developed a new geospatial database of paddy rice agriculture for 13 countries in South and Southeast Asia. These countries have 鈭悸30% of the world population and 鈭悸2/3 of the total rice land area in the world. We used 8-day composite images (500-m spatial resolution) in 2002 from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor onboard the NASA EOS Terra satellite. Paddy rice fields are characterized by an initial period of flooding and transplanting, during which period a mixture of surface water and rice seedlings exists. We applied a paddy rice mapping algorithm that uses a time series of MODIS-derived vegetation indices to identify the initial period of flooding and transplanting in paddy rice fields, based on the increased surface moisture. The resultant MODIS-derived paddy rice map was compared to national agricultural statistical data at national and subnational levels. Area estimates of paddy rice were highly correlated at the national level and positively correlated at the subnational levels, although the agreement at the national level was much stronger. Discrepancies in rice area between the MODIS-derived and statistical datasets in some countries can be largely attributed to: (1) the statistical dataset is a sown area estimate (includes multiple cropping practices); (2) failure of the 500-m resolution MODIS-based algorithm in identifying small patches of paddy rice fields, primarily in areas where topography restricts field sizes; and (3) contamination by cloud. While further testing is needed, these results demonstrate the potential of the MODIS-based algorithm to generate updated datasets of paddy rice agriculture on a timely basis. The resultant geospatial database on the area and spatial distribution of paddy rice is useful for irrigation, food security, and trace gas emission estimates in those countries.

DOI

[3]
Kuenzer C, Knauer K.Remote sensing of rice crop areas[J]. International Journal of Remote Sensing, 2013,34(6):2101-2139.Rice means life for millions of people and it is planted in many regions of the world. It primarily grows in the major river deltas of Asia and Southeast Asia, such as the Mekong Delta, known as the Rice Bowl of Vietnam, the second-largest rice-producing nation on Earth. However, Latin America, the USA, and Australia have extensive rice-growing regions. In addition, rice is the most rapidly growing source of food in Africa. Rice is therefore of significant importance to food security in an increasing number of low-income food-deficit countries. This review article gives a complementary overview of how remote sensing can support the assessment of paddy rice cultivation worldwide. This article presents and discusses methods for rice mapping and monitoring, differentiating between the results achievable using different sensors of various spectral characteristics and spatial resolution. The remote sensing of rice-growing areas can not only contribute to the precise mapping of rice areas and the assessment of the dynamics in rice-growing regions, but can also contribute to harvest prediction modelling, the analyses of plant diseases, the assessment of rice-based greenhouse gas (methane) emission due to vegetation submersion, the investigation of erosion-control-adapted agricultural systems, and the assessment of ecosystem services in rice-growing areas.

DOI

[4]
Gumma M K, Thenkabail P S, Maunahan A, et al.Mapping seasonal rice cropland extent and area in the high cropping intensity environment of Bangladesh using MODIS 500m data for the year 2010[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2014,91:98-113.

[5]
Hill M J, Donald G E.Estimating spatio-temporal patterns of agricultural productivity in fragmented landscapes using AVHRR NDVI time series[J]. Remote Sensing of Environment, 2003,84(3):367-384.The characteristics of Normalized Difference Vegetation Index (NDVI) time series can be disaggregated into a set of quantitative metrics that may be used to derive information about vegetation phenology and land cover. In this paper, we examine the patterns observed in metrics calculated for a time series of 8 years over the southwest of Western Australia-an important crop and animal production area of Australia. Four analytical approaches were used; calculation of temporal mean and standard deviation layers for selected metrics showing significant spatial variability; classification based on temporal and spatial patterns of key NDVI metrics; metrics were analyzed for eight areas typical of climatic and production systems across the agricultural zone; and relationships between total production and productivity measured by dry sheep equivalents were developed with time integrated NDVI (TINDVI). Two metrics showed clear spatial patterns; the season duration based on the smooth curve produced seven zones based on increasing length of growing season; and TINDVI provided a set of classes characterized by differences in overall magnitude of response, and differences in response in particular years. Frequency histograms of TINDVI could be grouped on the basis of a simple shape classification: tall and narrow with high, medium or low mean indicating most land is responsive agricultural cover with uniform seasonal conditions; broad and short indicating that land is of mixed cover type or seasonal conditions are not spatially uniform. TINDVI showed a relationship to agricultural productivity that is dependent on the extent to which crop or total agricultural production was directly reduced by rainfall deficiency. TINDVI proved most sensitive to crop productivity for Statistical Local Areas (SLAs) having rainfall less than 600 mm, and in years when rainfall and crop production were highly correlated. It is concluded that metrics from standardized NDVI time series could be routinely and transparently used for retrospective assessment of seasonal conditions and changes in vegetation responses and cover.

DOI

[6]
Atzberger C, Rembold F.Mapping the spatial distribution of winter crops at sub-pixel level using AVHRR NDVI time series and neural nets[J].Remote Sensing, 2013,5(3):1335-1354.For large areas, it is difficult to assess the spatial distribution and inter-annual variation of crop acreages through field surveys. Such information, however, is of great value for governments, land managers, planning authorities, commodity traders and environmental scientists. Time series of coarse resolution imagery offer the advantage of global coverage at low costs, and are therefore suitable for large-scale crop type mapping. Due to their coarse spatial resolution, however, the problem of mixed pixels has to be addressed. Traditional hard classification approaches cannot be applied because of sub-pixel heterogeneity. We evaluate neural networks as a modeling tool for sub-pixel crop acreage estimation. The proposed methodology is based on the assumption that different cover type proportions within coarse pixels prompt changes in time profiles of remotely sensed vegetation indices like the Normalized Difference Vegetation Index (NDVI). Neural networks can learn the relation between temporal NDVI signatures and the sought crop acreage information. This learning step permits a non-linear unmixing of the temporal information provided by coarse resolution satellite sensors. For assessing the feasibility and accuracy of the approach, a study region in central Italy (Tuscany) was selected. The task consisted of mapping the spatial distribution of winter crops abundances within 1 km AVHRR pixels between 1988 and 2001. Reference crop acreage information for network training and validation was derived from high resolution Thematic Mapper/Enhanced Thematic Mapper (TM/ETM+) images and official agricultural statistics. Encouraging results were obtained demonstrating the potential of the proposed approach. For example, the spatial distribution of winter crop acreage at sub-pixel level was mapped with a cross-validated coefficient of determination of 0.8 with respect to the reference information from high resolution imagery. For the eight years for which reference information was available, the root mean squared error (RMSE) of winter crop acreage at sub-pixel level was 10%. When combined with current and future sensors, such as MODIS and Sentinel-3, the unmixing of AVHRR data can help in the building of an extended time series of crop distributions and cropping patterns dating back to the 80s.

DOI

[7]
Potgieter A B, Apan A,Dunn P, et al.Estimating crop area using seasonal time series of enhanced vegetation index from MODIS satellite imager[J]. Australian Journal of Agricultural Research, 2007,58(4):316-325.Cereal grain is one of the main export commodities of Australian agriculture. Over the past decade, crop yield forecasts for wheat and sorghum have shown appreciable utility for industry planning at shire, state, and national scales. There is now an increasing drive from industry for more accurate and cost-effective crop production forecasts. In order to generate production estimates, accurate crop area estimates are needed by the end of the cropping season. Multivariate methods for analysing remotely sensed Enhanced Vegetation Index (EVI) from 16-day Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery within the cropping period (i.e. April-November) were investigated to estimate crop area for wheat, barley, chickpea, and total winter cropped area for a case study region in NE Australia. Each pixel classification method was trained on ground truth data collected from the study region. Three approaches to pixel classification were examined: (i) cluster analysis of trajectories of EVI values from consecutive multi-date imagery during the crop growth period; (ii) harmonic analysis of the time series (HANTS) of the EVI values; and (iii) principal component analysis (PCA) of the time series of EVI values. Images classified using these three approaches were compared with each other, and with a classification based on the single MODIS image taken at peak EVI. Imagery for the 2003 and 2004 seasons was used to assess the ability of the methods to determine wheat, barley, chickpea, and total cropped area estimates. The accuracy at pixel scale was determined by the percent correct classification metric by contrasting all pixel scale samples with independent pixel observations. At a shire level, aggregated total crop area estimates were compared with surveyed estimates. All multi-temporal methods showed significant overall capability to estimate total winter crop area. There was high accuracy at pixel scale (> 98% correct classification) for identifying overall winter cropping. However, discrimination among crops was less accurate. Although the use of single-date EVI data produced high accuracy for estimates of wheat area at shire scale, the result contradicted the poor pixel-scale accuracy associated with this approach, due to fortuitous compensating errors. Further studies are needed to extrapolate the multi-temporal approaches to other geographical areas and to improve the lead time for deriving cropped-area estimates before harvest.

DOI

[8]
Zheng B J, Campbell J B, de Beurs K M. Remote sensing of crop residue cover using multi-temporal Landsat imagery[J]. Remote Sensing of Environment, 2012,177:99-183.Tillage practices, which have direct impacts on soil and water quality, have changed dramatically during the past several decades. Tillage information is one of the important inputs for environmental modeling, but the availability of this information is still limited spatially and temporally. Previous studies have encountered difficulties in defining reliable correlations between crop residue cover (CRC) and Landsat-based tillage indices because they neglected the significance of the timing of tillage implementation. This study explores relationships between temporal changes of agricultural surfaces and the normalized difference tillage index (NDTI) in Central Indiana. We found that minimum NDTI (minNDTI) values extracted from multi-temporal NDTI profiles reliably indicate the surface status when tillage or planting occurred. Simple linear regression reveals a coefficient of determination (R 2 ) of 0.89 between CRC and minNDTI for calibration. In addition, a percentage change (PC) method was tested for classifying CRC into three categories (CRC02<0230%; 30%02<02CRC02<0270%; CRC02>0270%). Both the minNDTI and PC methods resulted in overall classification accuracies of >0290%, producer's accuracies of 83–100%, and user's accuracies of 75–100%. Our results indicated that both Landsat TM and ETM+ imagery are capable of mapping CRC, however, multi-temporal Landsat imagery is required. To establish a capability for crop residue mapping, designers of future remote sensing platforms should consider increasing temporal resolution.

DOI

[9]
Jia K, Wu B F, Li Q Z.Crop classification using HJ satellite multispectral data in the north China plain[J]. Journal of Applied Remote Sensing, 2013,7(1):287-297.The HJ satellite constellation is designed for environment and disaster monitoring by the Chinese government. This paper investigates the performance of multitemporal multispectral charge-coupled device (CCD) data on board HJ-1-A and HJ-1-B for crop classification in the North China Plain. Support vector machine classifier is selected for the classification using different combinations of multitemporal HJ multispectral data. The results indicate that multitemporal HJ CCD data could effectively identify wheat fields with an overall classification accuracy of 91.7%. Considering only single temporal data, 88.2% is the best classification accuracy achieved using the data acquired at the flowering time of wheat. The performance of the combination of two temporal data acquired at the jointing and flowering times of wheat is almost as well as using all three temporal data, indicating that two appropriate temporal data are enough for wheat classification, and much more data have little effect on improving the classification accuracy. Moreover, two temporal data acquired over a larger time interval achieves better results than that over a smaller interval. However, the field borders and smaller cotton fields cannot be identified effectively by HJ multispectral data, and misclassification phenomenon exists because of the relatively coarse spatial resolution.

DOI

[10]
Pan Y Z, Li L, Zhang J S, et al.Winter wheat area estimation from MODIS-EVI time series data using the crop proportion phenology index[J] . Remote Sensing of Environment, 2012,119:232-242.The global distribution of croplands is of critical interest to a wide group of end-users. Different crops have their own representative phenological stages during their growing seasons, which differ considerably from other natural vegetation types. During the last decade, the Moderate Resolution Imaging Spectroradiometer (MODIS) has become a key tool for vegetation monitoring because of its high temporal resolution, extensive scope, and rapid availability of various products. However, mixed pixels caused by the moderate spatial resolution produce significant errors in crop area estimation. Here we propose a Crop Proportion Phenology Index (CPPI) to express the quantitative relationship between the MODIS vegetation index (VI) time series and winter wheat crop area. The utility of this index was tested in two experimental areas in China: one around Tongzhou and the other around Shuyang, as representative districts around a metropolis and a rural area, respectively. The CPPI performed well in these two regions, with the root mean square error (RMSE) in fractional crop area predictions ranging roughly from 15% in the individual pixels to 5% above 6.25聽km 2 . The training samples containing mixtures of crop types mitigated the challenges of pure end-member selection in a spectral mixture analysis. A small number of training samples are sufficient to generate the CPPI, which is adaptable to other crop types and larger regions. Estimating the CPPI parameters across larger spatial scales helped improve the stability of the model.

DOI

[11]
Wardlow B D, Egbert S L, Kastens J H.Analysis of time-series MODIS 250m vegetation index data for crop classification in the U.S. central great plains[J]. Remote Sensing of Environment, 2007,108(3):290-310.The global environmental change research community requires improved and up-to-date land use/land cover (LULC) datasets at regional to global scales to support a variety of science and policy applications. Considerable strides have been made to improve large-area LULC datasets, but little emphasis has been placed on thematically detailed crop mapping, despite the considerable influence of management activities in the cropland sector on various environmental processes and the economy. Time-series MODIS 250m Vegetation Index (VI) datasets hold considerable promise for large-area crop mapping in an agriculturally intensive region such as the U.S. Central Great Plains, given their global coverage, intermediate spatial resolution, high temporal resolution (16-day composite period), and cost-free status. However, the specific spectral鈥搕emporal information contained in these data has yet to be thoroughly explored and their applicability for large-area crop-related LULC classification is relatively unknown. The objective of this research was to investigate the general applicability of the time-series MODIS 250m Enhanced Vegetation Index (EVI) and Normalized Difference Vegetation Index (NDVI) datasets for crop-related LULC classification in this region. A combination of graphical and statistical analyses were performed on a 12-month time-series of MODIS EVI and NDVI data from more than 2000 cropped field sites across the U.S. state of Kansas. Both MODIS VI datasets were found to have sufficient spatial, spectral, and temporal resolutions to detect unique multi-temporal signatures for each of the region's major crop types (alfalfa, corn, sorghum, soybeans, and winter wheat) and management practices (double crop, fallow, and irrigation). Each crop's multi-temporal VI signature was consistent with its general phenological characteristics and most crop classes were spectrally separable at some point during the growing season. Regional intra-class VI signature variations were found for some crops across Kansas that reflected the state's climate and planting time differences. The multi-temporal EVI and NDVI data tracked similar seasonal responses for all crops and were highly correlated across the growing season. However, differences between EVI and NDVI responses were most pronounced during the senescence phase of the growing season.

DOI

[12]
Ward B D, Egbert S L.Large-area crop mapping using time-series MODIS 250m NDVI data: an assessment for the U.S. Central great plains[J]. Remote Sensing of Environment, 2008,112(3):1096-1116.Improved and up-to-date land use/land cover (LULC) data sets that classify specific crop types and associated land use practices are needed over intensively cropped regions such as the U.S. Central Great Plains, to support science and policy applications focused on understanding the role and response of the agricultural sector to environmental change issues. The Moderate Resolution Imaging Spectroradiometer (MODIS) holds considerable promise for detailed, large-area crop-related LULC mapping in this region given its global coverage, unique combination of spatial, spectral, and temporal resolutions, and the cost-free status of its data. The objective of this research was to evaluate the applicability of time-series MODIS 250m normalized difference vegetation index (NDVI) data for large-area crop-related LULC mapping over the U.S. Central Great Plains. A hierarchical crop mapping protocol, which applied a decision tree classifier to multi-temporal NDVI data collected over the growing season, was tested for the state of Kansas. The hierarchical classification approach produced a series of four crop-related LULC maps that progressively classified: 1) crop/non-crop, 2) general crop types (alfalfa, summer crops, winter wheat, and fallow), 3) specific summer crop types (corn, sorghum, and soybeans), and 4) irrigated/non-irrigated crops. A series of quantitative and qualitative assessments were made at the state and sub-state levels to evaluate the overall map quality and highlight areas of misclassification for each map.The series of MODIS NDVI-derived crop maps generally had classification accuracies greater than 80%. Overall accuracies ranged from 94% for the general crop map to 84% for the summer crop map. The state-level crop patterns classified in the maps were consistent with the general cropping patterns across Kansas. The classified crop areas were usually within 1&ndash;5% of the USDA reported crop area for most classes. Sub-state comparisons found the areal discrepancies for most classes to be relatively minor throughout the state. In eastern Kansas, some small cropland areas could not be resolved at MODIS' 250m resolution and led to an underclassification of cropland in the crop/non-crop map, which was propagated to the subsequent crop classifications. Notable regional areal differences in crop area were also found for a few selected crop classes and locations that were related to climate factors (i.e., omission of marginal, dryland cropped areas and the underclassification of irrigated crops in western Kansas), localized precipitation patterns (overclassification of irrigated crops in northeast Kansas), and specific cropping practices (double cropping in southeast Kansas).

DOI

[13]
Zhong L H, Gong P, Biging G S.Efficient corn and soybean mapping with temporal extendibility: a multi-year experiment using Landsat Imagery[J]. Remote Sensing of Environment, 2014,140:1-13.

[14]
Dong J W, Xiao X M, Kou W L, et al.Tracking the dynamics of paddy rice planting area in 1986-2010 through time series Landsat images and phenology-based algorithms[J]. Remote Sensing of Environment, 2015,160:99-113.

[15]
骆剑承,周成虎,沈占锋,等.遥感信息图谱计算的理论方法研究[J].地球信息科学学报,2009,11(5):664-669.遥感应用的本质是投入专业知识从对地观测影像中提炼专题信息,并以之服务于各类分析与决策的过程。其中,信息计算是整个遥感应用服务技术链的基底。本文在传承地学信息图谱理论的基础上,提出了遥感&quot;图-谱&quot;信息耦合的空间认知理论,构建了&quot;像元&mdash;基元&mdash;目标&mdash;格局&quot;为一体的遥感信息图谱计算的理论方法体系,将其分为&quot;像元级&quot;和&quot;对象级&quot;两个层次,并阐述了高性能计算环境支持下,遥感信息图谱计算平台的设计开发思路及目前研发进展,总结了遥感信息图谱计算的发展趋势和重点研究问题。

DOI

[ Luo J C, Zhou C H, Shen Z F, et al.Theoretic and methodological review on sensor information tupu computation[J].Journal of Geo-information Science, 2009,11(5):664-669. ]

[16]
周伟,关键,姜涛,等.多光谱遥感影像中云影区域的检测与修复[J].遥感学报,2012,16(1):137-142.提出了一种有效针对多光谱遥感影像的云影检测与阴影区域修复方法。基于同一地区时相相近的两幅影像，充分利用碎云及阴影的光谱特性分别对云影区域进行融合增强，然后采用Otsu算法求解最佳阈值自动检测出云及阴影区域，根据云影的出现会引起两幅影像局部相应区域明显的亮度变化，可排除亮地物和水体的影响，建立归一化的云影密度图，在此基础上，采用线性加权组合与光谱直方图匹配相结合的方法对其加以修复，利用SPOT 4影像进行的实验表明其修复效果完全能够满足应用需要。

[ Zhou W, Guan J, Jiang T, et al.Automaticdetection and repairing of cloud and shadow regions in multi-spectral remote sensing images[J]. Journal of Remote Sensing, 2012,16(1):132-142. ]

[17]
Steven M D, Malthus T J, Baret F, et al.Inter-calibration of vegetation indices from different sensor systems[J].Remote Sensing of Environment, 2003,88(12):412-422.

[18]
张宏斌,杨桂霞,李刚,等.基于MODIS NDVI和NOAA NDVI数据的空间尺度转换方法研究——以内蒙古草原区为例[J].草业科学,2009,26(10):39-45.<p>选择内蒙古自治区的主要草原植被类型区作为研究区域，分析了2000-2003年生长季(4-10月)内月NOAA NDVI和MODIS NDVI数据之间相关性的年季变化规律。采用统计学方法，探讨了在大尺度空间范围内进行不同分辨率遥感数据之间的空间尺度转换方法，并利用2002年7、8月的NOAA NDVI和MODIS NDVI数据建立NOAA NDVI空间尺度转化模型，对该模型进行了时间尺度外推和精度验证。结果表明：针对NOAA NDVI和MODIS NDVI数据建立的空间尺度转化模型具有较好的应用效果，能够用于时间尺度外推，使MODIS和NOAA NDVI数据之间的分析比较具有科学性和有效性。</p>

[ Zhang H B, Yang G X, Li G.Study on the MODIS NDVI and NOAA NDVI based spatial scaling method-a case study in Inner Mongolia[J].Pratacaltural Science, 2009,26(10):39-45. ]

[19]
张焕雪,曹新,李强子,等.基于多时相环境星NDVI时间序列的农作物分类研究[J].遥感技术与应用,2015,30(2):304-311.<p>时相和归一化植被指数(NDVI)时间序列特征在农作物分类提取方面具有重要的应用价值。以黑龙江红星农场为研究区,利用多时相环境星HJ\|1 A/B CCD数据及其多期平滑重构后的NDVI时间序列曲线特征,在对象尺度上采用决策树算法开展了农作物分类研究,通过与单独利用多时相遥感数据分类结果的对比分析,研究了增加NDVI时序曲线特征对分类精度的影响。结果表明:面向对象分类方法得到的地块较为规则,平滑了地块内部同种作物间的噪声,避免了&ldquo;椒盐现象&rdquo;适合于我国东北地区农作物分类识别;利用NDVI时序曲线特征参与分类,增强了不同作物之间的光谱差异,提高了作物分类精度,比仅使用3个多时相HJ\|1 A/B CCD数据分类精度提高了5.45%,Kappa系数提高了0.09。通过该研究探讨了NDVI时序曲线特征在作物分类中的应用,拓展了遥感数据在农业领域的应用范围,具有推广价值。</p>

DOI

[ Zhang H X, Cao X, Li Q Z, et al.Research on crop identification using multi-temporal NDVI HJ images[J]. Remote Sensing Technology and Application, 2015,30(2):304-311. ]

[20]
张峰,吴炳方,刘成林,等.利用时序植被指数监测作物物候的方法研究[J].农业工程学报,2004,20(1):155-159.该文是对全国主要产粮县旱地和水田作物的物候期进行遥感监测。在数据预处理中采用最小二乘法和谐函数分解重构方法相结合，去除时序植被指数影像的云噪声影响。基于土地利用数据，通过耕地植被指数加权平均的方法提取区旱地和水田作物生长过程。结合野外观察数据，对一年一熟作物用作物生长过程的最大上升斜率、最大值和最大下降斜率作为作物出苗(返青)期、抽穗期和收获期的遥感识别标志。对一年两熟、多熟县作物物候期依据轮作规律进行了监测。同时进行物候年际间对比和农业灾害监测分析。遥感监测出苗(返青)期和收获期与野外采样照片实测信息有90%的相同率，抽穗期遥感监测与实测信息相同率95%。

DOI

[ Zhang F, Wu B F, Liu C L, et al.Methods of monitoring crop phonological stages using time series of vegetation indicator[J].Transactions of the Chinese Society of Agricultural Engineering, 2004,20(1):155-159. ]

[21]
Jonsson P, Eklundh L.Seasonalityextraction by function fitting to time-series of satellite sensor data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2002,40(8):1824-1832.Not Available

DOI

[22]
李治,杨晓梅,梦樊,等.物候特征辅助下的随机森林宏观尺度土地覆盖分类方法研究[J].遥感信息,2013,28(6):48-55.通过遥感技术获取大范围土地覆 盖信息对于监测、理解和预测自然资源具有重要的科学意义。MODIS数据是当今宏观尺度土地覆盖研究的主要数据源。本文以河北省为研究区,应用 MOD13Q1数据产品,构建MODIS NDVI时间序列,从中反演物候特征作为参与分类的主要辅助信息,并采用随机森林分类方法进行宏观尺度土地覆被分类实验,并与单决策树(CART)进行对 比分析。实验结果表明,物候特征辅助下的随机森林宏观尺度土地覆被分类方法的总体精度为87.2%,Kappa系数为0.83,比CART单一决策树精度 提高了17.9%;应用物候特征参与分类,使得总体精度提高2.6%;其中,旱地和建筑用地精度分别提高了6.7%和11.9%。

DOI

[ Li Z, Yang X M, Meng F, et al.LULC classification based on random forest with aid of phonological features[J]. Remote Sensing Information, 2013,28(6):48-55. ]

Options
Outlines

/

〈 〉