农作物种植面积遥感估算的影响因素研究

引用本文

张焕雪, 李强子, 文宁, 杜鑫, 陶青山, 田亦陈. 农作物种植面积遥感估算的影响因素研究[J]. 国土资源遥感, 2015,27(4): 54-61
ZHANG Huanxue, LI Qiangzi, WEN Ning, Du Xin, TAO Qingshan, TIAN Yichen. Important factors affecting crop acreage estimation based on remote sensing image classification technique[J]. REMOTE SENSING FOR LAND & RESOURCES,2015,27(4): 54-61 复制到剪切板

Permissions

《国土资源遥感》编辑部

农作物种植面积遥感估算的影响因素研究

张焕雪¹, 李强子¹, 文宁², 杜鑫¹, 陶青山², 田亦陈¹

1.中国科学院遥感与数字地球研究所,北京 100101

2.湖南省国土资源规划院,长沙 410007

通信作者:李强子(1970-),男,博士,研究员,主要从事农业、生态与高分辨率遥感监测技术研究。Email: liqz@radi.ac.cn。

第一作者简介:张焕雪(1988-),女,博士研究生,主要研究方向为农业遥感。Email: zhanghx@radi.ac.cn。

收稿日期: 2014-07-16

基金项目: 国家自然科学基金项目“作物种植面积遥感估算不确定性研究”(编号: 41071277)、中国科学院遥感与数字地球研究所所长青年基金(编号: Y3SJ7000CX)、湖南省国土资源厅基础测绘项目(编号: 2012-8-1)和“十二五”国家科技支撑计划项目“大麻遥感监测技术与毒性诊断仪研制”(编号: 2011BAK04B01)”共同资助

摘要

针对不同的农作物种植结构区,研究影响遥感影像分类各因素与农作物种植面积估算精度的定性和定量关系是十分必要的。以RapidEye影像提取的早稻种植信息为研究对象,从农作物的种植成数、种植破碎度和地块形状指数3个角度进行了不同空间分辨率下各因素对农作物面积监测的影响研究。结果表明:随着农作物种植成数的降低,种植结构越来越破碎,种植地块趋于狭长分布,各分辨率下农作物面积估算精度均呈递减趋势;要达到85%以上的面积估算精度,当作物种植成数在50%以上时,可选取高于150 m分辨率的遥感数据;当作物种植较为破碎时,需要在提高影像空间分辨率的同时融入其他技术手段;当作物种植地块为狭长分布时,提高影像的空间分辨率并不能保证面积估算精度,必须通过其他技术手段达到精度要求;并最终得到了4种影响因素对面积估算精度的定量评估模型。研究结果为解决不同农作物种植结构区遥感数据的选择、面积估算精度的提高,以及在特定研究区和数据源条件下可达到的面积估算水平等问题提供了理论基础。

关键词: 遥感估算; 种植面积; 农作物; 影像分类; 精度

中图分类号:TP 751.1; P 237 文献标志码:A 文章编号:1001-070X(2015)04-0054-08 doi: 10.6046/gtzyyg.2015.04.09

Important factors affecting crop acreage estimation based on remote sensing image classification technique

ZHANG Huanxue¹, LI Qiangzi¹, WEN Ning², Du Xin¹, TAO Qingshan², TIAN Yichen¹

1. Institute of Remote Sensing and Digital Earth, Beijing 100101, China

2. Hunan Province Land and Resources Planning Institute, Changsha 410007, China

Abstract

It is necessary and valuable to study the effect of influencing factors of crop classification on crop acreage estimation from both qualitative and quantitative points of view. Therefore, the authors analyzed the resolution effect on the acreage estimation accuracy by using RapidEye imagery. Spatial statistics methods and manifold accuracy evaluation indices were used respectively to analyze the data with different index statistics of crop proportion, crop fragmentation and shape. The results show that decreased crop proportion and increased crop fragmentation and shape index will lead to reducing regional accuracy under all resolutions. And in order to keep the accuracy higher than 85%, we can select any resolution higher than 150 m data when the crop proportion is higher than 50%, so as to achieve the accuracy requirements. As merely improving resolution cannot guarantee the crop acreage estimation accuracy when the crop land exhibits long and narrow distribution, other technology must be adopted in this case. Finally the quantitative influence model of the four factors for crop acreage estimation accuracy is built. The results of this paper would provide academic reference for resolving the problem of data selection and accuracy improvement in crop acreage estimation by remote sensing.

Keyword: remote sensing estimation; planting acreage; crop; image classification; accuracy

Show Figures

0 引言

农作物种植面积是影响农产品产量的重要因素之一^[1]。快速、准确、可靠的农作物种植面积调查结果, 已经成为国家农业政策分析和粮食宏观决策的重要支持信息之一^[2]。遥感技术因其具有大范围的宏观观测能力和客观性, 一直在农作物种植面积调查中发挥着重要作用, 利用遥感影像进行各种农作物识别与面积估算的技术和方法也得到快速发展。农作物类型遥感识别从最初的人工目视判读^[3]到利用影像光谱特征进行监督和非监督分类^{[1, 4, 5]}, 目前已经在利用遥感影像的时相特征^{[6, 7, 8]}、散射极化特征^{[9, 10, 11]}、空间特征^{[12, 13]}等领域取得全面突破。为了提高分类精度, 还发展了混合像元分解技术、面向对象的影像分类技术、多源数据融合技术^{[14, 15]}、辅助数据分析技术^[16]以及新的分类识别模式^{[17, 18]}等。

目前, 农作物的遥感识别已经形成了多源、多尺度遥感数据在像元、亚像元和对象尺度的各类监督和非监督分类方法, 并发展了各类参数化和非参数化的分类算法^[19]。但更高的分类精度是不断追求的目标。尽管科学家们充分利用遥感技术获取多源遥感数据, 并一直致力于提高农作物的分类精度^{[14, 20]}; 但目前关于不同影像特征(分辨率、时相等)以及农作物种植地块特征等对分类精度的影响方面, 还没有得出一致性结论。针对某一具体研究区, 要满足预期的分类精度, 应当设计什么样的监测方案, 选用多大分辨率的遥感影像, 采用哪些特征进行分类, 确定的监测方案和数据源可以达到多高的分类精度等等, 一直没有给出满意的结果^[21]。

本文以湖南省澧县水稻产区为实验区, 以利用5 m分辨率的RapidEye影像提取的早稻种植信息为研究对象, 模拟一系列不同分辨率的遥感数据, 旨在分析影响农作物遥感识别的各主要因素(种植成数、种植破碎度和地块形状指数)在不同分辨率下对农作物面积估算精度的影响规律, 建立各因素对农作物面积估算精度的综合影响模型, 为解决不同作物种植结构区遥感数据的选择、面积估算精度的提高方式, 以及特定研究区和数据源可以达到的面积估算精度水平等问题提供理论依据和实验基础。

1 实验区概况与数据源

1.1 实验区概况

本文的实验区位于湖南省中北部澧县和临澧县境内, 中心位置的地理坐标为E111° 39'21″, N29° 49'37″(图1)。

	Figure Option View Download New Window
	图1 实验区地理位置及RapidEye影像Fig.1 Geo-location and RapidEye image of experimental area

该区属亚热带湿润气候区, 年降水量为1 200~1 500 mm; 四季温差明显, 年平均气温16° ~18℃。种植的农作物类型主要有水稻、棉花和少量的大棚葡萄、蔬菜等。其中, 水稻是主要的粮食作物, 其种植结构主要是双季稻或单季稻。单季中稻主要是6月初移栽, 9月初收获; 双季早稻生长季主要是从5月初到7月中下旬; 晚稻主要是7月下旬种植, 11月收获。

1.2 数据源

根据实验区主要农作物的物候历及遥感影像的质量, 本文选取了2012年6月19日获取的RapidEye影像。该影像清晰度好, 覆盖实验区面积约为25 km× 20 km(图1); 空间分辨率5 m, 幅宽77 km, 5个波段的波谱范围分别是B1(0.44~0.51 μ m), B2(0.52~0.59 μ m), B3(0.63~0.685 μ m), B4(0.69~0.73 μ m)和B5(0.76~0.85 μ m), 其中B4是红边波段。成像时双季早稻处于抽穗期, 而单季中稻处于移栽期, 比较适合早稻信息的提取。

对RapidEye影像的处理主要包括辐射定标和几何纠正。辐射定标^[17]是将影像的DN值转化为大气顶归一化光谱反射率, 并采用FLAASH模型^[22]进行大气校正; 几何纠正则以USGS提供的实验区TM影像作为参考影像, 采用二次多项式方法对RapidEye数据进行几何精纠正, 选取了40个地面控制点, 使纠正误差控制在0.5个像元内。

此外, 在早稻提取过程中, 还采用了取自国土资源二调土地利用数据集中的湖南省水田分布数据以及2012年6月采集的7个野外样方数据(图2)。

	Figure Option View Download New Window
	图2 水田及7个地面调查样方Fig.2 Paddy field and 7 field survey plots

2 研究方法

2.1 技术路线

本文首先利用RapidEye 数据分类进行实验区早稻种植地块的提取, 作为实验研究的近似真值; 然后以该早稻提取结果为基础, 通过尺度扩展构建不同分辨率影像序列, 并将实验区均匀布设成1 km× 1 km样区, 计算每个样区内早稻的种植成数、种植破碎度及地块形状指数指标; 再通过统计面积估算精度及标准差, 分析作物分类的主要影响因素(种植成数、种植破碎度和地块形状指数)随不同影像空间分辨率对农作物面积估算精度的影响规律; 最终建立影像空间分辨率和上述3个指标共4种因素对农作物面积估算精度的综合影响模型。总体技术路线如图3所示。

	Figure Option View Download New Window
	图3 各因素影响分析技术流程Fig.3 Flowchart of all factors analysis

2.2 早稻种植地块提取

在实验区1:1万比例尺水田数据的支持下, 首先从RapidEye影像中提取出水田区域; 然后利用实地调查数据, 选取8 319个像元作为训练样本, 采用最大似然分类(MLC)、支持向量机(SVM)和人工神经网络(ANN)3种分类方法得到早稻种植地块分布信息, 并估算其面积。根据野外实测样方得到的7 142个验证样本对分类结果进行精度验证, 将精度最高的SVM分类结果(总体分类精度95.99%, Kappa=0.89)作为5 m分辨率早稻分布的“ 准真值” , 以便进行后续研究。

2.3 多分辨率影像序列获取

本文通过尺度扩展方法对利用RapidEye影像提取的早稻种植地块分布数据进行重采样, 获得不同分辨率的影像序列。常用的尺度扩展方法主要包括简单平均法、中心像元法、最近邻法、双线性内插和立方卷积等^{[23, 24]}。其中, 最近邻法、双线性内插和立方卷积是常用的方法; 但Hay等^[25]发现, 这3种方法在尺度转换因子大于5时, 不适合将影像从高分辨率转换到低分辨率。故本文采用了简单平均法进行影像的尺度扩展。

2.4 遥感分类精度的影响因素

通过将实验中分类得到的农作物类型与地面调查得到的实地农作物分布类型进行比较分析, 发现农作物的种植成数^[26](即面积比例)、种植破碎度^[27](即集聚程度)以及地块的形状指数^[28]均对农作物遥感识别精度具有明显的影响。因此, 本文将25 km× 20 km的实验区均分成500个面积为1 km× 1 km的样区, 计算每个样区内原始5 m分辨率和一系列较低分辨率下的早稻的种植成数、种植破碎度和地块形状指数; 分析其尺度效应, 并通过多元线性回归分析得到4种因素与面积估算精度之间的关系模型。2.5 精度评价为全面分析上述4种因素对农作物面积监测的影响, 本文从面积估算精度和估算精度的稳定性2个方面进行评价。

1)面积估算精度K_i是用来比较相对精度的指标, 按式(1)计算得到, 即

$\begin{matrix} K_{i} = (1 - \frac{│ A_{i} - A_{0} │}{A_{0}}) \times 100 % \end{matrix}$ , (1)

式中:A₀为从5 m分辨率RapidEye影像得到的早稻面积, 本文将该值作为基准值; A_i为由尺度转换得到的某种较低分辨率(i)下提取的早稻面积。

2)估算精度的稳定性以面积估算精度平均值的标准差δ ₍_i₎来衡量, 用来反映整个实验区各个样区间早稻估算精度的离散程度(即与整个实验区早稻平均面积估算精度 $\begin{matrix} {\bar{K}}_{i} \end{matrix}$ 的偏离程度), 计算方法为

$\begin{matrix} δ_{(i)} = \frac{1}{n} \overset{n}{\sum_{i = 1}} │ K_{i} - {\bar{K}}_{i} │^{2} \end{matrix}$ 。 (2)

3 结果与分析

3.1 种植成数的影响

本文通过计算不同分辨率下的早稻种植面积百分比, 发现随着种植成数的升高, 面积估算精度呈增加趋势, 而且精确度也逐渐趋于稳定。图4给出了10 m分辨率模拟影像中500个1 km× 1 km样区的面积估算精度与种植成数的二维散点图。

	Figure Option View Download New Window
	图4 10m分辨率影像分类时农作物种植成数与面积估算精度关系散点图Fig.4 Scatter plots of crop proportion and area estimation accuracy under 10 m resulation image classification

从图4可以看出, 在较低的种植成数水平下, 面积估算精度具有较大的分散性(即精确度较低); 而随着种植成数的提高, 面积估算精度趋于收敛, 表明作物面积估算误差也越来越小; 且能实现对作物面积的无偏估计, 当种植成数超过60%, 面积估算精度稳定在85%以上。

为进一步考察随影像空间分辨率的变化作物种植成数与面积估算精度的关系, 本文将种植成数指标从0~100%进行分组统计, 共分为10组(组间距10%), 分别统计不同空间分辨率下、不同种植成数水平与对应样区的面积估算精度和平均标准差的变化趋势(图5)。

	Figure Option View Download New Window
	图5 不同分辨率影像分类时农作物种植成数与面积估算精度(左)/标准差(右)的关系Fig.5 Relationship between crop proportion and area estimation accuracy (left)/standard deviation(right) under different resolution image classifications

从图5可以看出, 随着农作物种植成数的增加, 各分辨率下的作物面积估算精度均呈增加趋势, 而标准差均呈递减趋势。作物种植成数低于10%时, 各分辨率下的面积估算精度均处于最低值, 而标准差均处于最大值; 且随着空间分辨率的降低, 面积估算精度越来越低、标准差值越来越大, 即面积估算精度的离散程度越来越大, 面积估算结果越来越不稳定。当分辨率降至150 m时, 面积估算精度仅为45%, 标准差则大于25%。表明在种植成数低于10%时, 要达到85%以上的面积估算精度, 且保证估算标准差接近于0, 需要采用分辨率高于10 m的遥感数据源。随着种植成数和影像分辨率的提高, 各分辨率下的面积估算精度呈现出较快的增长趋势, 并在种植成数达到40%~50%时接近最高并逐渐稳定在85%以上; 相反, 面积估算精度标准差则随着种植成数和分辨率的提高出现显著的降低趋势, 并在达到40%~50%的种植成数水平时, 稳定在4%以下。说明当作物种植百分比在50%以上时, 要想达到85%的分类精度, 可以选取分辨率高于150 m的遥感数据。

3.2 种植破碎度的影响

由于我国自然环境和种植制度的复杂性, 作物“ 插花” 种植现象普遍存在^[29]。本文研究了农作物种植破碎度与种植面积估算精度的关系, 并对500个样区的结果进行了统计分析, 发现随着种植破碎度的提高, 种植面积估算精度趋于降低(图6)。

	Figure Option View Download New Window
	图6 10m分辨率影像分类时农作物种植破碎度与面积估算精度关系散点图Fig.6 Scatter plots of crop fragmentation and area estimation accuracy under 10 m resulation image classification

从图6可以看出, 在农作物种植破碎度较高的情况下, 面积估算精度具有趋于分散和低精确度特征。亦即随着种植破碎度的降低, 面积估算精度趋于收敛且稳定; 而随着作物种植的分散化, 面积估算误差不断递增, 当种植破碎度高于0.7时, 面积估算精度很难保证足够的精度, 基本在85%以下。

为进一步考察不同空间分辨率影像分类中作物种植破碎度与面积估算精度的关系, 对种植破碎度指标0~1.3进行分组实验, 共分9组(组间距如图7所示), 分别统计不同的种植破碎度级别对应样区的面积估算精度和平均标准差在不同空间分辨率下的变化趋势(图7)。

	Figure Option View Download New Window
	图7 不同分辨率影像分类时农作物种植破碎度与面积估算精度(左)/标准差(右)的关系Fig.7 Relationship between crop fragmentation and area estimation accuracy (left)/standard deviation(right) under different resulation image classifications

从图7可以看出, 随着种植破碎度的增加和影像分辨率的降低, 各种分辨率条件下的作物面积估算精度均呈递减趋势, 标准差则基本呈递增趋势。作物种植破碎度指数为0~0.5时, 种植较为完整, 各分辨率下的面积估算精度均处于最高水平, 标准差均处于最低水平; 随着影像空间分辨率的降低, 面积估算精度越来越低, 当分辨率降低至150 m时, 面积估算精度仅为55%。表明在0~0.5破碎度等级下, 要达到85%以上的面积估算精度, 且保证估算标准差接近于0, 需要采用空间分辨率高于20 m的遥感数据。随着种植破碎度指数从0.5逐步增加, 各分辨率下的面积估算精度呈降低趋势, 标准差表现出增加趋势; 但分辨率为10 m的数据除外, 该分辨率下作物面积估算标准差随种植破碎度的变化基本不变, 说明当采用高于10 m分辨率的遥感数据时, 作物种植破碎度已经不是影响其面积估算稳定性的主要因素。当种植破碎度达到1.2~1.3(即作物分布极为破碎)时, 即使采用10 m分辨率的数据, 面积估算精度也难以达到60%。说明当作物种植较为破碎时, 要想达到85%的面积估算精度, 不仅需要采用高于10 m空间分辨率的遥感影像, 而且还需要通过其他技术手段提高面积估算精度。

3.3 地块形状指数的影响

受地形地势及灌溉条件等限制, 不同地区的农作物地块形状各不相同。本文通过计算地块形状指数并分析其与面积估算精度的关系, 发现随着形状指数的提高(即地块由正方形向细长形过度), 面积估算精度明显下降(图8)。

	Figure Option View Download New Window
	图8 10m分辨率影像分类时地块形状指数与面积估算精度关系散点图Fig.8 Scatter plots of shape index and area estimation accuracy under 10 m resulation image classification

从图8可以看出, 当作物种植多为狭长地块时, 面积估算精度较低; 随着地块形状的变长, 面积估算精度递减且分布趋于分散。当地块形状指数低于0.13时, 面积估算精度才能基本稳定在85%以上。

为进一步考察随影像空间分辨率的变化地块形状指数与面积估算精度的关系, 本文对形状指数0~0.26进行分组实验, 共分15组(组间距如图9所示), 分别统计不同的地块形状指数层次在不同分辨率影像分类后对面积估算精度和平均标准差的影响。

	Figure Option View Download New Window
	图9 不同分辨率影像分类时地块形状指数与面积估算精度(左)/标准差(右)的关系Fig.9 Relationship between shape index and area estimation accuracy (left)/standard deviation(right) under different resulation image classifications

从图9可以看出, 随着作物种植地块趋向于狭长分布和影像分辨率的降低, 各分辨率下作物面积估算精度均呈递减趋势, 标准差则基本呈增加趋势。农作物地块形状指数小于0.1时, 各分辨率下的面积估算精度均处于最高水平, 标准差均处于最低水平; 随着影像空间分辨率的降低, 面积估算精度越来越低。当分辨率降低至150 m时, 面积估算精度降到25%以下。即使选用分辨率为10 m的数据, 面积估算精度也仅为70%。表明在地块形状指数小于0.1的情况下, 要达到85%以上的估算精度, 且保证估算标准差接近于0, 需要采用分辨率高于10 m的遥感数据。随着地块形状指数从0.1逐步增加(即种植地块逐步趋向于狭长分布), 各分辨率下的面积估算精度均呈降低趋势, 同时标准差表现出增加趋势; 尤其是当地块形状指数达到0.25~0.26时, 在不同分辨率下面积估算精度均不足30%。说明当作物种植地块呈狭长形状的情况下, 需要采用分辨率高于10 m(甚至5 m)的遥感影像来提高其面积估算精度。

3.4 多因素综合影响分析与建模

根据3.1— 3.3节中得到的影像空间分辨率、种植成数、种植破碎度和地块形状指数4种农作物面积估算影响因素分析资料, 进行多元回归统计拟合, 建立了4种因素对面积估算精度的综合影响模型。

建模过程中的估计参数如表1所示。拟合优度R²值越大, 所反映的自变量与因变量的共变量比率越高, 模型与数据的拟合程度越好。

表1 分类模型的估计参数 Tab.1 Estimate parameters of classification model

分类模型的回归分析结果(表2)表明, 本模型的显著性水平sig.< 0.05, 可以认为所建立的回归方程有效。

表2 分类模型的回归分析参数 Tab.2 Regression analysis parameters of classification model

多元统计模型的相关参数及模型的显著性检验结果见表3。

表3 多元统计模型参数 Tab.3 Parameters of multivariate statistical model

多元统计模型中4个自变量回归系数的显著性水平Sig.=0.000均小于0.05, 可以认为4个自变量(种植成数、地块形状指数、种植破碎度和影像空间分辨率)对因变量(面积估算精度)均有显著影响。最终, 回归分析得到的回归方程为

Y=98.70+0.73X₁+4.83X₂-4.78X₃-0.23X₄, (3)

式中:Y为面积估算精度; X₁为种植成数; X₂为地块形状指数; X₃为种植破碎度; X₄为影像空间分辨率。

4 结论与展望

4.1 结论

本文通过简单平均法将原始空间分辨率为5 m的遥感影像分别模拟到10 m, 20 m, 30 m, 60 m, 90 m, 120 m和150 m分辨率, 其中10 m, 20 m, 30 m和60 m分辨率分别对应农作物种植面积遥感测量中常用的SPOT, CBERS, TM和IRS-P6数据。通过本文研究, 最终得到了影响农作物分类精度的主要因素(种植成数、种植破碎度、地块形状指数)随不同影像分辨率的变化对农作物种植面积估算精度的影响规律, 以及4种因素对农作物面积估算精度的综合影响模型。为解决不同研究区数据源的选取、农作物面积估算精度的提高方式以及特定研究区和数据源可以达到的农作物面积估算精度水平等问题提供了理论基础。得出如下结论:

1)随着农作物种植成数的增加, 各影像分辨率下作物面积估算精度均呈增加趋势, 标准差呈递减趋势。作物种植成数百分比为0~10%时, 要达到85%以上的面积估算精度, 且保证估算标准差接近于0, 需要采用分辨率高于10 m的遥感数据源。当作物种植成数百分比在50%以上时, 在不同影像分辨率下面积估算精度基本均稳定在88%以上, 标准差稳定在4%以下; 要想达到85%的分类精度, 可以选取分辨率高于150 m的任意遥感数据。

2)随着农作物种植越来越破碎, 各影像分辨率下作物面积估算精度均呈递减趋势, 标准差基本呈增加趋势。作物种植破碎度指数为0~0.5时, 作物种植较为完整, 要达到85%以上的面积估算精度, 且保证估算标准差接近于0, 需要采用分辨率高于20 m的遥感数据源。当作物种植较为破碎(例如破碎度指标为1.2~1.3)时, 要想达到85%的面积估算精度, 除了需要提高影像的空间分辨率, 还需要通过多元回归统计拟合等其他技术手段提高面积估算精度。

3)随着农作物种植地块逐步趋向于狭长分布, 各影像分辨率下作物面积估算精度均呈递减趋势, 标准差呈增加趋势。农作物地块形状指数为0~0.1时, 要达到85%以上的面积估算精度, 且保证估算标准差接近于0, 需要采用分辨率高于10 m的遥感数据源。当农作物种植地块为狭长分布(例如形状指数指标为0.25~0.26)时, 仅仅提高影像的空间分辨率并不能保证面积估算精度, 还需要通过其他技术手段才能达到85%的面积估算精度需求。

4)4种影响因素(种植成数X₁、地块形状指数X₂、种植破碎度X₃和影像空间分辨率X₄)对农作物面积估算的精度均有显著影响, 具体的定量影响方式为:Y=98.70+0.73X₁+4.83X₂-4.78X₃-0.23X₄。

4.2 展望

在以后的研究中还需在以下方面做进一步讨论和改进:

1)本文所采用的一系列较低空间分辨率数据, 是由5m分辨率的RapidEye多光谱影像中的早稻提取信息基于简单平均法模拟得到的, 这样做是为了避免其他因素的影响而单纯地分析遥感影像空间分辨率对农作物面积估算精度的影响。在今后的研究中将尝试直接采用不同空间分辨率影像进行分类精度评价, 并对这2种不同的方法进行比较分析, 以期得到更加符合实际情况的结论。

2)本文设计的农作物提取结果只区分了目标作物和非目标作物2类。在今后的研究中, 将考虑对更多的其他作物进行更为全面的影响因素分析, 并尝试在其他不同地貌类型和种植结构的典型作物区进行试验, 以进一步测试本文结论的普适性。

The authors have declared that no competing interests exist.

参考文献

View Option

[1]	吴炳方. 中国农情遥感速报系统[J]. 遥感学报, 2004, 8(6): 481-497. Wu B F. China crop watch system with remote sensing[J]. Journal of Remote Sensing, 2004, 8(6): 481-497. [本文引用:2]
[2]	Chen J. Rapid urbanization in China: A real challenge to soil protection and food security[J]. Catena, 2007, 69(1): 1-15. [本文引用:1]
[3]	蒋雪中, 赵锐, 李强子, 等. 利用GVG线采样技术提取农作物种植面积及其精度分析[J]. 南京气象学院学报, 2002, 25(1): 78-83. Jiang X Z, Zhao R, Li Q Z, et al. Extraction of crop acreage using GVG system and its precision analysis[J]. Journal of Nanjing Institute of Meteorology, 2002, 25(1): 78-83. [本文引用:1]
[4]	梁友嘉, 徐中民. 基于SPOT-5卫星影像的灌区作物识别[J]. 草业科学, 2013, 30(2): 161-167. Liang Y J, Xu Z M. Crop identification in the irrigation district based on SPOT-5 satellite imagery[J]. Pratacultural Science, 2013, 30(2): 161-167. [本文引用:1]
[5]	刘磊, 江东, 徐敏, 等. 基于多光谱影像和专家决策法的作物分类研究[J]. 安徽农业科学, 2011, 39(25): 15809-15811. Liu L, Jiang D, Xu M, et al. Crops classification based on multi-spectral image and decision tree method[J]. Journal of Anhui Agricultural Sciences, 2011, 39(25): 15809-15811. [本文引用:1]
[6]	欧文浩, 苏伟, 薛文振, 等. 基于HJ-1卫星影像的三大农作物估产最佳时相选择[J]. 农业工程学报, 2010, 26(11): 176-182. Ou W H, Su W, Xue W Z, et al. Selection of optimum phase for yield estimation of three major crops based on HJ-1 satellite images[J]. Transactions of the Chinese Society of Agricultural Engineering, 2010, 26(11): 176-182. [本文引用:1]
[7]	Wardlow B D, Egbert S L. Large-area crop mapping using time-series MODIS 250 m NDVI data: An assessment for the U. S. Central Great Plains[J]. Remote Sensing of Environment, 2008, 112(33): 1096-1116. [本文引用:1]
[8]	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): 073576. [本文引用:1]
[9]	Yonezawa C, Negishi M, Azuma K, et al. Growth monitoring and classification of rice fields using multitemporal RADARSAT-2 full-polarimetric data[J]. International Journal of Remote Sensing, 2012, 33(18): 5696-5711. [本文引用:1]
[10]	Jia K, Li Q Z, Tian Y C, et al. Crop classification using multi-configuration SAR data in the North China Plain[J]. International Journal of Remote Sensing, 2012, 33(1): 170-183. [本文引用:1]
[11]	马红章, 刘素美, 朱晓波, 等. 基于被动微波遥感技术的玉米冠层叶面积指数反演[J]. 国土资源遥感, 2013, 25(3): 66-71. doi: DOI: 106046/gtzyyg. 2013. 03. 12. Ma H Z, Liu S M, Zhu X B, et al. Crop LAI inversion based on the passive microwave remote sensing technology[J]. Remote Sensing for Land and Resources, 2013, 25(3): 66-71. doi: DOI:10.6046/gtzyyg.2013.03.12. [本文引用:1]
[12]	Balaguer A, Ruiz L A, Hermosill T, et al. Definition of a comprehensive set of texture semivariogram features and their evaluation for object-oriented image classification[J]. Computers and Geosciences, 2010, 36(2): 231-240. [本文引用:1]
[13]	Berberoglu S, Curran P J, Lloyd C D, et al. Texture classification of Mediterranean land cover[J]. International Journal of Applied Earth Observation and Geoinformation, 2007, 9(3): 322-334. [本文引用:1]
[14]	Li Q Z, Wu B F, Jia K, et al. Maize acreage estimation using ENVISAT MERIS and CBERS-02B CCD data in the North China Plain[J]. Computers and Electronics in Agriculture, 2011, 78(2): 208-214. [本文引用:2]
[15]	胥兵, 方臣. ZY-10C星图像与ETM+图像融合方法及效果评价[J]. 国土资源遥感, 2014, 26(3): 80-85. doi: DOI: 106046/gtzyyg. 2014. 03. 13. Xu B, Fang C. Data fusion methods of ZY-10C and ETM+ images and effect evaluation[J]. Remote Sensing for Land and Resources, 2014, 26(3): 80-85. doi: DOI:10.6046/gtzyyg.2014.03.13. [本文引用:1]
[16]	程乾, 王人潮. 数字高程模型和多时相MODIS数据复合的水稻种植面积遥感估算方法研究[J]. 农业工程学报, 2005, 21(5): 89-92. Cheng Q, Wang R C. Estimation of the rice planting area using digital elevation model and multitemporal moderate resolution imaging spectroradiometer[J]. Transactions of the Chinese Society of Agricultural Engineering, 2005, 21(5): 89-92. [本文引用:1]
[17]	杨小唤, 张香平, 江东. 基于MODIS时序NDVI特征值提取多作物播种面积的方法[J]. 资源科学, 2004, 26(6): 17-22. Yang X H, Zhang X P, Jiang D. Extraction of multi-crop planting areas from MODIS data[J]. Resources Science, 2004, 26(6): 17-22. [本文引用:2]
[18]	于菲菲, 曾永年, 徐艳艳, 等. 基于植被分区的多特征遥感智能分类[J]. 国土资源遥感, 2014, 26(1): 63-70. doi: DOI: 106046/gtzyyg. 2014. 01. 12. Yu F F, Zeng Y N, Xu Y Y, et al. Intelligent remote sensing classification of multi-character data based on vegetation partition[J]. Remote Sensing for Land and Resources, 2014, 26(1): 63-70. doi: DOI:10.6046/gtzyyg.2014.01.12. [本文引用:1]
[19]	权文婷, 王钊. 冬小麦种植面积遥感提取方法研究[J]. 国土资源遥感, 2013, 25(4): 8-15. doi: DOI: 106046/gtzyyg. 2013. 04. 02. Quan W T, Wang Z. Researches on the extraction of winter wheat planting area using remote sensing method[J]. Remote Sensing for Land and Resources, 2013, 25(4): 8-15. doi: DOI:10.6046/gtzyyg.2013.04.02. [本文引用:1]
[20]	贾坤, 李强子, 田亦陈, 等. 微波后向散射数据改进农作物光谱分类精度研究[J]. 光谱学与光谱分析, 2011, 31(2): 483-487. Jia K, Li Q Z, Tian Y C, et al. Accuracy improvement of spectral classification of crop using microwave backscatter data[J]. Spectroscopy and Spectral Analysis, 2011, 31(2): 483-487. [本文引用:1]
[21]	张焕雪, 李强子. 空间分辨率对作物识别及种植面积估算的影响研究[J]. 遥感信息, 2014, 29(2): 38-42. Zhang H X, Li Q Z. Effects of spatial resolution on crop identification and acreage estimation[J]. Remote Sensing Information, 2014, 29(2): 38-42. [本文引用:1]
[22]	Vermote E F, Tanré D, Deuzé J L, et al. Second simulation of the satellite signal in the solar spectrum, 6S: An overview[J]. IEEE Transactions on Geoscience and Remote Sensing, 1997, 35(3): 675-686. [本文引用:1]
[23]	Kavzoglu T. Simulating land sat ETM+ imagery using DAIS 7915 hyperspectral scanner data[J]. International Journal of Remote Sensing, 2004, 25(22): 5049-5067. [本文引用:1]
[24]	Woodcock C, Harward V J. Nested-hierarchical scene models and image segmentation[J]. International Journal of Remote Sensing, 1992, 13(16): 3167-3187. [本文引用:1]
[25]	Hay G J, Niemann K O, Goodenough D G. Spatial thresholds, image-objects, and upscaling: A multiscale evaluation[J]. Remote Sensing of Environment, 1997, 62(1): 1-19. [本文引用:1]
[26]	李强子, 吴炳方, 许文波. 农作物分类成数的精度检验[J]. 遥感学报, 2004, 8(6): 588-597. Li Q Z, Wu B F, Xu W B. Accuracy assessment of crop type proportion using GVG instrument on transect line[J]. Journal of Remote Sensing, 2004, 8(6): 588-597. [本文引用:1]
[27]	胡潭高, 张锦水, 潘耀忠, 等. 景观破碎度在冬小麦面积抽样设计中的应用研究[J]. 遥感学报, 2010, 14(6): 1117-1138. Hu T G, Zhang J S, Pan Y Z, et al. Application of land scape fragmentation in winter wheat area sampling design[J]. Journal of Remote Sensing, 2010, 14(6): 1117-1138. [本文引用:1]
[28]	尹锴, 赵千钧, 文美平, 等. 海岛型城市森林景观格局效应及其生态系统服务评估[J]. 国土资源遥感, 2014, 26(2): 128-133. doi: DOI: 106046/gtzyyg. 2014. 02. 21. Yin K, Zhao Q J, Wen M P, et al. Assessment of land scape pattern effect and ecosystem services of island urban forest[J]. Remote Sensing for Land and Resources, 2014, 26(2): 128-133. doi: DOI:10.6046/gtzyyg.2014.02.21. [本文引用:1]
[29]	吴炳方, 李强子. 基于两个独立抽样框架的农作物种植面积遥感估算方法[J]. 遥感学报, 2004, 8(6): 551-569. Wu B F, Li Q Z. Crop acreage estimation using two individual sampling frameworks with stratification[J]. Journal of Remote Sensing, 2004, 8(6): 551-569. [本文引用:1]

2004

0.0

吴炳方. 中国农情遥感速报系统[J]. 遥感学报, 2004, 8(6): 481-497.

Wu B

. China crop watch system with remote sensing[J]. Journal of Remote Sensing, 2004, 8(6): 481-497.

The paper introduces China Crop Watch System with Remote Sensing (CropWatch),including its contents, technical innovation and operational development, validation and future activities. Crop Watch contents 5 components, consisting of crop condition monitoring, crop acreage estimation and yield prediction, grain production estimation, cropping index and crop structure monitoring, and drought monitoring. Each component is described briefly in this paper and detailed in following papers. Since 1998, the CropWatch has been put in operation. Each year, there are 7 monthly bulletins released, covering crop condition, crop acreage and yield, drought, agro-meteorology, cropping index and crop structure, and also grain production. And there are 21 10-days announcements released, covering drought and crop condition. During the operational period since 1998, the technology used by the CropWatch are consistently improved. All AVHRR image from 1991 are processed for calibration and correction precisely which began in 2000, including sensor calibration, zenith correction, cloudy detection and masking, geometric correction, atmospheric correction, BRDF correction,and NDVI and surface temperature calculation. The CropWatch uses both different image and growing profile of NDVI to monitor the crop condition, estimates crop acreage using remote sensing and sampling technique based on stratification. The cluster-sample framework is used to estimate plant proportion with TM and Radarsat image as samples, and transect-sampling framework is used to survey crop type proportion in planted-area. At the same time, The Crop Watch continuously expands its monitoring area. From 1998 to 2003, its monitoring area has been expanded from east China to national wide, to north and south America, to Australia and Thailand. Meanwhile, the Crop Watch continuously expands its monitoring contents, from crop condition monitoring, crop acreage estimation and crop yield prediction, to grain production, crop structure, and cropping index. The Crop Watch has its own validation programme. Each year, a few validation sites have been selected to do the field measurement for verifying specific results. For example, in 1999, Jiangning in South China were Dehui in North China were two sites. The crop LAI every 10 days and crop type proportion have been measured to verify the crop condition and crop acreage. Validation shows that the accuracy are up to 97% for plant proportion and 95 crop type proportion.

介绍了中国农情遥感速报系统的建设情况 ,系统内容包括农作物长势监测、农作物种植面积监测、农作物单产预测与粮食产量估算、作物时空结构监测和粮食供需平衡预警等。简要介绍了 1998年以来中国农情遥感速报系统在监测内容与监测范围、监测频率、技术发展以及质量控制与过程检验体系建立等方面的进展 ,并就中国农情遥感速报系统的发展方向提出了展望。

... 0 引言农作物种植面积是影响农产品产量的重要因素之一^[1] ...

... 农作物类型遥感识别从最初的人工目视判读^[3]到利用影像光谱特征进行监督和非监督分类^[1,4,5],目前已经在利用遥感影像的时相特征^[6,7,8]、散射极化特征^[9,10,11]、空间特征^[12,13]等领域取得全面突破 ...

2007

0.0

... 快速、准确、可靠的农作物种植面积调查结果,已经成为国家农业政策分析和粮食宏观决策的重要支持信息之一^[2] ...

2002

0.0

2013

0.0

梁友嘉, 徐中民. 基于SPOT-5卫星影像的灌区作物识别[J]. 草业科学, 2013, 30(2): 161-167.

Liang Y

, Xu Z

. Crop identification in the irrigation district based on SPOT-5 satellite imagery[J]. Pratacultural Science, 2013, 30(2): 161-167.

High resolution satellite imagery is one of important data sources for crop monitoring and assessment, and has important application values in the field of crop planning and yield estimation. Distribution of crops in the Yingke irrigation district of Zhangye City was analyzed by using a combined data of SPOT5 image, obtained in 2008, with four spectral bands (green, red, nearinfrared and shortwave infrared) and 2.5 m pixel size covering. Two images with pixel sizes of 10 m and 30 m were also generated from the original combined image to simulate coarser resolution satellite imagery. Five supervised classification techniques, including minimum distance, Mahalanobis distance, maximum likelihood, spectral angle mapper (SAM) and support vector machine (SVM), were applied to identify crop types in the study. The effects of pixel size on classification results were also examined. Kappa analysis showed that the maximum likelihood and SVM, though there were no statistical differences between them, performed better than those from other classified methods and the value of Kappa were 0.871 9 and 0.862 5, respectively. Accuracy assessment showed that the maximum likelihood gave the best result with overall accuracy values of 90.6%. The results also showed that increasing pixel size from 2.5 m to 10 m or 30 m did not significantly affect the classification accuracy for crop identification. Overall results indicate that SPOT5 image in conjunction with maximum likelihood and SVM classification techniques can be used for identifying crop types and estimating crop areas.

高分辨率卫星影像是作物精确分类和评估的重要数据源，在农作物种植规划、估产等领域具有重要的应用价值。本研究利用分辨率为2.5 m的SPOT5影像分析张掖市盈科灌区的作物分布状况，同时分别生成分辨率为10和30 m的影像，用于尺度验证。最终得到研究区作物分类图，所用方法主要有最小距离法、马氏距离、最大似然法、光谱角制图仪（SAM）和支持向量机（SVM）。Kappa系数分析表明，最大似然法和SVM的分类效果好于其它分类器，分别为0.871 9和0.862 5，但这两种方法的统计量无明显区别；分类图精度评价表明，基于最大似然法的分类图总体精度最高，为90.6%；随像元空间尺度的增加，分类精度未产生明显变化。研究结果表明，最大似然法和SVM技术可以与SPOT5影像结合，用于作物类型识别和作物面积估算。

2011

0.0

2010

0.0

欧文浩, 苏伟, 薛文振, 等. 基于HJ-1卫星影像的三大农作物估产最佳时相选择[J]. 农业工程学报, 2010, 26(11): 176-182.

Ou W

, Su

, Xue W

, et al. Selection of optimum phase for yield estimation of three major crops based on HJ-1 satellite images[J]. Transactions of the Chinese Society of Agricultural Engineering, 2010, 26(11): 176-182.

For precisely estimating crop yield by remote sensing data, selection of the optimum phase is one of the key links.?In this study, China HJ-1 satellite CCD images were used for the estimation of three major crops (rice, corn, soybean) in 852 farm of Heilongjiang Province. Through wavelet transform filter and NDVI time series data of auto regressive moving average method, the optimum phase of estimation was established based on smoothed NDVI time series curves of 852 farm. Results showed that the fitting error for rice was -0.003356508, and it was -0.001687117 for corn and -0.002530646 for soybean, respectively. The smoothed NDVI time series curves match the local phenology ideally, and the optimum phase for yield estimation of three major crops can be determined by fitting curves.

对于农作物遥感估产，精确选择最佳估产时相是关键环节。该文利用中国自行研发的HJ-1卫星CCD影像对黑龙江八五二农场3大作物（水稻、玉米、大豆）进行遥感估产的最佳时相选择，通过构建小波变换滤波方法和移动平均法的时序NDVI曲线数据，并依据平滑后的时序NDVI曲线分别确定3大作物的遥感估产最佳时相。研究结果表明，从平滑后的时序NDVI曲线中识别出来的3大作物的关键生长期与当地作物的物候期相对比，水稻生长期拟合误差为-0.003356508，玉米生长期拟合误差为-0.001687117，大豆生长期拟合误差为-0.002530646。拟合结果与当地物候期基本吻合，即可从拟合曲线中确定各作物的最佳遥感估产时相。

2008

0.0

2013

0.0

2012

0.0

2012

0.0

2013

0.0

马红章, 刘素美, 朱晓波, 等. 基于被动微波遥感技术的玉米冠层叶面积指数反演[J]. 国土资源遥感, 2013, 25(3): 66-71. doi: DOI: 106046/gtzyyg. 2013. 03. 12.

Ma H

, Liu S

, Zhu X

, et al. Crop LAI inversion based on the passive microwave remote sensing technology[J]. Remote Sensing for Land and Resources, 2013, 25(3): 66-71. doi: DOI:10.6046/gtzyyg.2013.03.12.

The multichannel dual-polarized radiation characteristic database of corn vegetation canopy was constructed by using the corn structure parameters field measured data and the Matrix-Doubling model, and the relations between the emissivity and the transmissivity of the corn canopy were obtained using regression analysis based on the database. These relationships were applied to the microwave radiation propagation equation to compute the microwave radiation bright temperature of the surface covered with corn canopy. The results show that the correlation of the measured LAI value and the model inversion LAI is higher than 0.9, which suggests that the multichannel passive microwave data have considerable application potential in the aspect of corn LAI inversion.

1. College of Science, China University of Petroleum, Qingdao 266580, China; 2. China Centre for Resource Satellite Data and Application, Beijing 100094, China; 3. College of Earth Science and Technology, China University of Petroleum, Qingdao 266580, China; 4. Geomatics College, Shandong University of Science and Technology, Qingdao 266590, China; 5. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing Applications, Chinese Academy of Sciences, Beijing 100101, China

基于玉米冠层结构参数实测数据和Matrix-Doubling(MD)模型构建了玉米出苗期至抽穗期的冠层多波段、双极化微波辐射特性模拟数据库; 通过对模拟数据的回归分析得到了玉米冠层在各波段的微波发射率及其与透过率之间的经验关系,并将经验关系应用于0阶微波辐射传输模型; 结合土壤发射率模型构建了玉米冠层覆盖地表的微波辐射亮温参数化计算模型,并基于该参数化模型、利用玉米样地微波亮温观测试验数据,采用迭代方法进行了玉米叶面积指数(LAI)的反演。研究表明,LAI反演值与实测值的相关系数 r >0.9,说明多波段被动微波遥感数据在植被冠层LAI反演方面具有较大的应用潜力。

2010

0.0

2007

0.0

2011

0.0

... 为了提高分类精度,还发展了混合像元分解技术、面向对象的影像分类技术、多源数据融合技术^[14,15]、辅助数据分析技术^[16]以及新的分类识别模式^[17,18]等 ...

... 尽管科学家们充分利用遥感技术获取多源遥感数据,并一直致力于提高农作物的分类精度^[14,20] ...

2014

0.0

胥兵, 方臣. ZY-10C星图像与ETM+图像融合方法及效果评价[J]. 国土资源遥感, 2014, 26(3): 80-85. doi: DOI: 106046/gtzyyg. 2014. 03. 13.

, Fang

. Data fusion methods of ZY-10C and ETM+ images and effect evaluation[J]. Remote Sensing for Land and Resources, 2014, 26(3): 80-85. doi: DOI:10.6046/gtzyyg.2014.03.13.

The successful launch of ZY-1 02C satellite has filled the gap in the high resolution remote sensing satellite developed by China itself. Based on the characteristics of the satellite image and data quality,the authors took ZY-1 02C P/MS sensor data of Russia's Siberia as an example to interpret the region's geological and mineral resources information for the purpose of analyzing the data processing and data quality of P/MS image in combination with the United States Landsat ETM + data of the region for data fusion. Four kinds of fusion methods, i.e., HSV,Brovery,Gramm-Schmidt and PCA fusion, were used for experimental comparison. The experimental results show that the image fused with Gramm-Schmidt method not only suffers from the least loss of spectral information but also has the best fusion effect,thus improving the spatial resolution of the fused image effectively. In other words, the precision of remote sensing geological interpretation is greatly improved by Gramm-Schmidt method. The results obtained by the authors can play an active role in further promoting the application of the domestic high resolution remote sensing satellites in China.

1. Hubei Geological Survey, Wuhan 430022, China; 2. College of Resources, China University of Geosciences, Wuhan 430074, China

资源一号（ZY-1）02C卫星的发射成功填补了我国自主研发高分辨率遥感卫星的空白。根据该卫星的图像特点和数据质量，以俄罗斯西伯利亚地区的ZY-1 02C卫星P/MS传感器数据为例，通过解译该地区的地质矿产信息，对P/MS图像数据的处理和质量情况进行分析；同时结合该地区的美国Landsat卫星ETM + 数据，将P/MS全色波段数据与ETM + 多光谱数据融合，并采用HSV，Brovery，Gram-Schmidt和PCA等4种融合方法进行实验对比。实验结果表明，用Gram-Schmidt法融合后的图像，光谱信息丢失最少，融合效果最好，可有效提高融合图像的空间分辨率，进而提高该地区的遥感地质解译精度。该研究成果将为进一步推进我国国产高分辨率遥感卫星的应用发挥积极作用。

2005

0.0

程乾, 王人潮. 数字高程模型和多时相MODIS数据复合的水稻种植面积遥感估算方法研究[J]. 农业工程学报, 2005, 21(5): 89-92.

Cheng

, Wang R

. Estimation of the rice planting area using digital elevation model and multitemporal moderate resolution imaging spectroradiometer[J]. Transactions of the Chinese Society of Agricultural Engineering, 2005, 21(5): 89-92.

Rice is the important staple crop in China, the estimation of rice yield is an important component of country, and remote sensing technology is an important means of estimation of rice yield. Because of the hilly region, more cloud and variety of plants, it is very difficult to estimate the rice yield. With the launching of TERRA, Moderate resolution Imaging Spectroradiometer (MODIS), with abundant information, quickly acquiring data and wide range of coverage, is new data for estimation of rice yield. This study considered the characteristics of rice planting, and the digital slope image derived from the digital elevation model map and multitemporal MODIS were used for the purpose of improving the classification accuracy of MODIS in large hilly region. The results show that the slope and mutitemporal MODIS image can improve the accuracy of estimation of rice area in comparison with only one MODIS image.

随着搭载在TERRA卫星上的中分辨率成像光谱仪(MODIS)的出现,它以数据丰富、时间分辨率高和覆盖范围广等特点，为水稻遥感估产提供了较好的数据源。该文利用水稻种植易受坡度影响的特性，从数字高程模型(Digital Elevation Models简称DEM)中提取坡度信息，考虑到MODIS能提供多时相及丰富的数据，采用DEM产生的坡度和两个时相MODIS影像数据及植被指数复合提取水稻种植面积，经过比较试验证明，在南方丘陵山区的复杂地形区域,多源信息复合相对于单纯利用单景影像数据可以明显提高水稻种植面积估算的精度。

2004

0.0

杨小唤, 张香平, 江东. 基于MODIS时序NDVI特征值提取多作物播种面积的方法[J]. 资源科学, 2004, 26(6): 17-22.

Yang X

, Zhang X

, Jiang

. Extraction of multi-crop planting areas from MODIS data[J]. Resources Science, 2004, 26(6): 17-22.

One of the major problems in crop monitoring study is to extract planting areas and their spatial distribution automatically. This paper presents an automatic approach to planting area extraction for mixed planting crops (winter wheat, spring maize, summer maize, and soybean) using moderate spatial resolution and high temporal resolution MODIS data around Beijing, China. With a spatial resolution of 250 meters, band 1 and band 2 of MODIS data in red and near-infrared spectral regions can be used for dynamic monitoring of crops. NDVI values increase with the growth of the crops, and gradually decrease after reaching the maximum at a certain growth stage of the crops. Because different crops have different growth stages, the NDVI peak values and their occurrences can be different. After investigating the planting structure of the main crops and analyzing the NDVI values of different crops from mid-March to mid-November of 2002 in Beijing, following results are obtained: a. NDVI value of winter wheat is higher than other objects in the end of March, and its maximum appears in the beginning of May; b. maximum value of spring maize NDVI appears in the beginning of August; c. maximum value of summer maize NDVI appears in the middle of August; d. maximum value of bean NDVI appears also in the beginning of August, but it can be recognized from spring maize by the growth stage. Spring maize is a single-season crop; whenas bean is a double-season crop. So the temporal changes of NDVI profiles were obtained. In corporation with 1:100,000 scale landuse data of Beijing, the planting areas of winter wheat, spring maize, summer maize and bean were extracted from MODIS data, and the total accuracy is over 95%. In this research, multiple-phase MODIS data were received during main growing seasons and preprocessed; NDVI temporal profiles of main crop types were generated; models for planting areas extraction were developed based on the analysis of temporal NDVI profiles; and spatial distribution map of planting areas of winter wheat, maize and soybean in Beijing in 2002 were created. The study suggests that it is possible to extract planting areas automatically from MODIS data for large areas. Accuracy analysis showed that the results are highly reliable, especially in plain areas.

MODIS数据的1、2波段是具有250米空间分辨率的红和近红外波段，并具有较高的时间分辨率，可对农作物进行动态跟踪监测。随着农作物的生长，NDVI值逐渐增大，并在一定生育期达到最大值后开始下降。由于不同作物具有不同的生育期，NDVI峰值的大小、出现的时间各不相同。通过对北京市主要农作物的种植结构调查和3月中旬到11月上旬的不同作物的NDVI值采样分析，得出：①冬小麦NDVI值3月下旬相对较高， 5月上旬最大；②春玉米的NDVI峰值出现于8月上旬；③夏玉米的NDVI峰值出现于8月中旬；④大豆的NDVI峰值也出现在8月中上旬，可通过物候历与春玉米区分开来（春玉米是单季作物，大豆是双季作物），通过峰值大小与夏玉米区分开来。结合北京市1：10万土地利用数据，通过NDVI值时序变化规律从MODIS数据中提取了冬小麦、春玉米、夏玉米、大豆等作物的种植面积，总体精度达到95%以上。

... 辐射定标^[17]是将影像的DN值转化为大气顶归一化光谱反射率,并采用FLAASH模型^[22]进行大气校正 ...

2014

0.0

于菲菲, 曾永年, 徐艳艳, 等. 基于植被分区的多特征遥感智能分类[J]. 国土资源遥感, 2014, 26(1): 63-70. doi: DOI: 106046/gtzyyg. 2014. 01. 12.

Yu F

, Zeng Y

, Xu Y

, et al. Intelligent remote sensing classification of multi-character data based on vegetation partition[J]. Remote Sensing for Land and Resources, 2014, 26(1): 63-70. doi: DOI:10.6046/gtzyyg.2014.01.12.

In order to effectively extract land use/land cover remote sensing information in a wide range of terrain complex area,the authors, taking the transition zone between Tibetan Plateau and the Loess Plateau in eastern Qinghai as the study area,studied the intelligent remote sensing classification of land use/land cover by using ant colony intelligent optimization algorithm(ACIOA)in this paper. Firstly,TM image,digital elevation model,slope and aspect data were selected as characteristic bands for classification. Secondly,the study area was divided into two parts using the normalized difference vegetation index(NDVI)so as to reduce the influence of different objects with the same spectrum. Finally,the classification rules were excavated using ACIOA,by which regional land use/cover information was extracted. The results show that the ACIOA classification of multi-character data based on vegetation partition is superior to the traditional remote sensing classification. The overall accuracy of the classification and the Kappa coefficient of ACIOA with multi-character data based on vegetation partition is 88.85% and 0.86 respectively. Therefore,this study provides an effective way for extracting land use/land cover information in large-area complex terrain.

1. School of Geosciences and Geomatics, Central South University, Changsha 410083, China; 2. Center for Geomatics and Sustainable Development Research, Central South University, Changsha 410083, China

为了有效地提取大范围地形复杂区域的土地利用/土地覆盖遥感信息，以位居青藏高原与黄土高原过渡地带的青海东部地区为研究区，研究基于蚁群智能优化算法（ant colony intelligent optimization algorithm，ACIOA）的土地利用/土地覆盖遥感智能分类。首先选用TM图像、DEM、坡度和坡向数据作为分类的特征波段；然后利用归一化植被指数NDVI对实验区数据进行植被分区；最后利用ACIOA算法进行分类规则挖掘，并依据分类规则进行土地利用/覆盖信息的提取。研究表明，基于植被分区的多特征蚁群智能分类的总体精度为88.85%， Kappa =0.86，优于传统的遥感图像分类方法，为大范围地形复杂区域的土地利用/土地覆盖遥感信息提取提供了有效的方法。

2013

0.0

权文婷, 王钊. 冬小麦种植面积遥感提取方法研究[J]. 国土资源遥感, 2013, 25(4): 8-15. doi: DOI: 106046/gtzyyg. 2013. 04. 02.

Quan W

, Wang

. Researches on the extraction of winter wheat planting area using remote sensing method[J]. Remote Sensing for Land and Resources, 2013, 25(4): 8-15. doi: DOI:10.6046/gtzyyg.2013.04.02.

Currently,the extraction of winter wheat area has become the key point of winter wheat growth monitoring and yield estimation. Therefore,how to improve the accuracy of winter wheat planting area retrieval is one of the important topics in agricultural remote sensing at present and in the future. The retrieval model should be varying with the spatial resolution of satellite image so as to accurately retrieve winter wheat planting area from remote sensing images,and this is one of the important points of remote sensing of winter wheat planting area. In this study,the winter wheat planting area retrieval models are grouped into 5 different types, i.e., visual interpretation and computer automatic classification,classification based on pixel,classification based on parcel,remote sensing interpretative model classification and classification based on texture. The main research methods were described and evaluated,and the problems existent in current research work and the future development of remote sensing method for estimation winter wheat planting area were discussed. The focusing of the attention on the estimation of winter wheat area in the district with complex terrain,high land fragmentation and complex planting structure,the combination of remote sensing data with GIS data and meteorological data, the integration of multi-source remote sensing data with multi-temporal remote sensing data,the combination of ground spectral measurement data with high spectral data,and the changing of the validation data constitute the study direction for improving the accuracy of winter wheat planting area retrieval.

Shaanxi Provincial Agricultural Remote Sensing Information Center, Xi'an 710014, China

提取冬小麦种植面积是开展冬小麦长势监测和估产工作的重要环节,如何提高其提取精度是国内外研究的热点。针对不同空间分辨率的遥感图像,采取不同的遥感解译模型,得到更高的信息提取精度,是利用遥感方法提取冬小麦种植面积的关键。在对国内、外调研的基础上,将冬小麦种植面积遥感提取研究方法归纳为目视解译与计算机自动分类、面向像元分类、基于地块分类、遥感解译模型分类和基于纹理特征分类等5类。综述评价了主要的冬小麦种植面积遥感提取研究方法,讨论了目前冬小麦种植面积遥感提取存在的问题及未来发展方向。关注地形复杂、耕地破碎度较高及种植结构复杂地区的冬小麦种植面积提取,遥感数据、GIS和气象数据等相结合、多源多时相遥感数据相结合、地面光谱测量数据与高光谱图像数据相结合,以及验证数据的改变等,是提高冬小麦种植面积遥感提取精度的研究方向。

... 目前,农作物的遥感识别已经形成了多源、多尺度遥感数据在像元、亚像元和对象尺度的各类监督和非监督分类方法,并发展了各类参数化和非参数化的分类算法^[19] ...

2011

0.0

贾坤, 李强子, 田亦陈, 等. 微波后向散射数据改进农作物光谱分类精度研究[J]. 光谱学与光谱分析, 2011, 31(2): 483-487.

Jia

, Li Q

, Tian Y

, et al. Accuracy improvement of spectral classification of crop using microwave backscatter data[J]. Spectroscopy and Spectral Analysis, 2011, 31(2): 483-487.

In the present study, VV polarization microwave backscatter data used for improving accuracies of spectral classification of crop is investigated. Classification accuracy using different classifiers based on the fusion data of Hi satellite multi-spectral and Envisat ASAR VV backscatter data are compared. The results indicate that fusion data can take full advantage of spectral information of HJ multi-spectral data and the structure sensitivity feature of ASAR VV polarization data. The fusion data enlarges the spectral difference among different classifications and improves crop classification accuracy. The classification accuracy using fusion data can be increased by 5 percent compared to the single Hi data. Furthermore, ASAR VV polarization data is sensitive to non-agrarian area of planted field, and VV polarization data joined classification can effectively distinguish the field border. VV polarization data associating with multi-spectral data used in crop classification enlarges the application of satellite data and has the potential of spread in the domain of agriculture.

[Jia Kun; Li Qiang-zi; Tian Yi-chen; Wu Bing-fang; Zhang Fei-fei; Meng Ji-hua] Chinese Acad Sci, Inst Remote Sensing Applicat, Beijing 100101, Peoples R China.;[Jia Kun] Chinese Acad Sci, Grad Univ, Beijing 100049, Peoples R China.

... 尽管科学家们充分利用遥感技术获取多源遥感数据,并一直致力于提高农作物的分类精度^[14,20] ...

2014

0.0

... 针对某一具体研究区,要满足预期的分类精度,应当设计什么样的监测方案,选用多大分辨率的遥感影像,采用哪些特征进行分类,确定的监测方案和数据源可以达到多高的分类精度等等,一直没有给出满意的结果^[21] ...

1997

0.0

... 辐射定标^[17]是将影像的DN值转化为大气顶归一化光谱反射率,并采用FLAASH模型^[22]进行大气校正 ...

2004

0.0

... 常用的尺度扩展方法主要包括简单平均法、中心像元法、最近邻法、双线性内插和立方卷积等^[23,24] ...

1992

0.0

... 常用的尺度扩展方法主要包括简单平均法、中心像元法、最近邻法、双线性内插和立方卷积等^[23,24] ...

1997

0.0

... 但Hay等^[25]发现,这3种方法在尺度转换因子大于5时,不适合将影像从高分辨率转换到低分辨率 ...

2004

0.0

李强子, 吴炳方, 许文波. 农作物分类成数的精度检验[J]. 遥感学报, 2004, 8(6): 588-597.

Li Q

, Wu B

, Xu W

. Accuracy assessment of crop type proportion using GVG instrument on transect line[J]. Journal of Remote Sensing, 2004, 8(6): 588-597.

In this paper,we selected a test area about 5km by 5km in Taigu of Shanxi Province to assess the accuracy of crop type proportion obtained with GVG instrument on transect line. Roads on the QuickBird imagery were digitized with length of 255km, and 16.4km of the roads were selected as transect line for field sampling.The sampling ratio was about 6.4%. Crop type proportion for each crops was obtained using GVG instrument on the transect line.323 pictures were taken with clear crop identification. The density of taking picture was about 1 picture for every 50 meter. At the same time, the intensive field works were taken to map the crop pattern in detail. QuickBird imagery was used to make parcel map with a unique identity code. Field survey recorded the crop type for each parcel. If more than one crop exist in a parcel,the proportions are assigned to different crops. The crop type proportion can be calcuated with detailed crop pattern map. By comparing the crop type proportion derived from sampling and mapping,we could draw conclusions; Crop type proportion from GVG instrument on transect line has a very high accuracy for main crops, 98.7% for maize, 98.98% for soybean and 95 12% for vegetables. The accuracy is acceptable for the operational use of China Crop Watch System. For other crop types with low proportion, the accuracy is low and cannot be used in China Crop Watch System. This means GVG instrument can only be used to obtain the main crop proportion. Crop type proportion from GVG instrument on transect line is somewhat larger than the true values, this means that the missing samples for some crop types with low proportion. In Taigu test area, sesame, cottage and some medicinal vegetables are not sampled on transect line with GVG instrument

选择山西太谷一个 5km× 5km的实验区 ,利用样条采样框架结合GVG农情采样系统调查农作物分类成数。同时借助QuickBird甚高分辨率遥感影像进行地面作物种植地块勾绘 ,并派出地面调查队伍进行作物填图 ,统计汇总出的农作物分类成数的真实值。然后将两种不同方法得出的分类成数进行对比 ,发现利用样条采样框架和GVG农情采样系统对于大宗粮食作物分类成数的调查相对误差在 3%以内 ,能够满足中国农情遥感速报系统的运行需要。而对于小成数作物的调查精度较低 ,且存在漏采现象 ,不能满足需求 ,同时也由于漏采现象的存在和图片判读的主观性。利用样条采样框架和GVG农情采样系统获取的大宗作物分类成数略大于真实值 ,存在少量的系统误差 ,需要进行地面验证并加以克服。

... 4 遥感分类精度的影响因素通过将实验中分类得到的农作物类型与地面调查得到的实地农作物分布类型进行比较分析,发现农作物的种植成数^[26](即面积比例)、种植破碎度^[27](即集聚程度)以及地块的形状指数^[28]均对农作物遥感识别精度具有明显的影响 ...

2010

0.0

胡潭高, 张锦水, 潘耀忠, 等. 景观破碎度在冬小麦面积抽样设计中的应用研究[J]. 遥感学报, 2010, 14(6): 1117-1138.

Hu T

, Zhang J

, Pan Y

, et al. Application of land scape fragmentation in winter wheat area sampling design[J]. Journal of Remote Sensing, 2010, 14(6): 1117-1138.

将破碎度指标引入到遥感抽样调查中,重点探讨了面积规模指标与破碎度指标在分层抽样中的适用性.实验结果表明:在种植结构破碎区域,面积规模指标与破碎度指标的相火系数均在0.7以上,可以作为分层标志,随着抽样格网不断增加,面积规模指标的相关系数不断增大,而破碎度指标的相关系数则不断减小.当格网小于100m×100m时,破碎度指标作为分层标志效果要优于面积规模指标;在种植结构规整区域,面积规模指标要始终优于破碎度指标,破碎度指标更适合于种植结构破碎、抽样单元较小的遥感抽样调查方案中作为分层标志进行应用.

2014

0.0

尹锴, 赵千钧, 文美平, 等. 海岛型城市森林景观格局效应及其生态系统服务评估[J]. 国土资源遥感, 2014, 26(2): 128-133. doi: DOI: 106046/gtzyyg. 2014. 02. 21.

Yin

, Zhao Q

, Wen M

, et al. Assessment of land scape pattern effect and ecosystem services of island urban forest[J]. Remote Sensing for Land and Resources, 2014, 26(2): 128-133. doi: DOI:10.6046/gtzyyg.2014.02.21.

Based on five Landsat TM/ETM + remote sensing images of Xiamen Island acquired from 1987 to 2007,the authors carried out the quantitative analysis and evaluation of urban landscape pattern effects on urban forest ecosystem services by using such methods as remote sensing interpretation of the landscape elements,landscape index,ecosystem services evaluation and gray relational analysis. Some conclusions have been reached: 1 The area of the forest in Xiamen Island had been reduced from 1987,and reached its lowest value in 2002; correspondingly,the value of ecosystem services reached its lowest value of 1.281 billion yuan per year in 2002. The value of ecosystem services increased to 1.363 billion yuan per year in 2007,of which the water conservation service value was the largest share,followed by carbon oxygen release service. 2 The results of gray correlation analysis show that the urban forest ecosystem service values in Xiamen Island have a remarkable correlation with the patch density,patch number and Shannon diversity index. The lower correlation coefficient with the built-up area indicates that the land used to built-up in Xiamen Island increased at the expense of the other landscape. The urban forest in the island was under better protection.

1. Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China; 2. Xiamen University of Technology, Xiamen 361024, China; 3. Key Lab of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen 361021, China

以海岛型城市厦门市为研究区，利用1987—2007年5个时期的TM/ETM + 遥感数据，综合运用景观要素遥感解译、景观指数法、生态服务价值量法及灰色关联分析方法，定量评价了景观格局效应对厦门城市森林生态系统服务的影响。研究结果表明：1厦门岛城市森林面积自1987年以来一直趋于减少，在2002年达到最低值；其生态服务价值同样在2002年达到最低水平，为12.81亿元/a；2007年有所上升，为13.63亿元/a，其中，占城市森林总生态服务价值比重最大的是涵养水源功能，其次为固碳释氧功能；2厦门岛城市森林生态服务价值与斑块密度、斑块数和香农多样性指数具有较高的关联度；与建成区面积的关联度较低，表明厦门岛建设用地以侵占城市森林以外的其他景观为主，岛内城市森林得到了较好的保护。

2004

0.0

吴炳方, 李强子. 基于两个独立抽样框架的农作物种植面积遥感估算方法[J]. 遥感学报, 2004, 8(6): 551-569.

Wu B

, Li Q

. Crop acreage estimation using two individual sampling frameworks with stratification[J]. Journal of Remote Sensing, 2004, 8(6): 551-569.

Crop plots are very small in China due to special farm-use rules.Image classification techniques are limited in crop acreage survey using remote sensing.In this paper,we analyze this problem and provide a substitute methodology to estimate crop acreage.In this methodology,crop statification is fundamentd.Proportional of main crop types as well as physical factors of temperature,precipitation,soil type and sun radiation are considered.There are about 11 strata in China at the first level based on physical factors,44 strata at the second level based on crop proportion and 102 strata at the third level based on arable land intensity.Two indicidual sampling frameworks are uesd.The cluster sampling is used to estimate the proportion of planted area on arable land with remote sensing data,mostly Landsat and Radarsat data,currently also ENVISAT ASAR data are used.The clusters are defined as a map sheet at a scale of 1:1:100 000 (about1/16 Landsat TM scene).And images are selected based on cluster sample randomly for each crop season.After atmospheric correction,geometric correction nonarable land masking,remotely sensed images are classified by ISODATA unsuoervised classification,and then the planted areas are labeled by considering NDVI value.The planted area proportions are calculated for each stratum.The transect sampling framework is used to estimate the proportions of different crop type within planted area.To identify the crop proportion of a small parcel,field works should be used since it is impossible to make crop classfication with remote sensing data cost-effectively.The transect sampling actually is a two-stage sampling.In the first stage,PSUs are selected randomly on a 4km*4km area frame.In the second stage,the selected PSUs are sampled only along the road within PSUs,called transect line.The sampling works in the field are to take pictures along the road within PSUs with 100m buffer,and a GVC system is designed for this purpose.Proportions of every crop type are calculated for each stratum.Crop acreages are calculated under the support of current arable database.For every crop type,the palnted acreage is the arable area multiply by planted proportion and crop type proportion of syratum.The estimation of early rice acreage in 2003 in China is presented as a case syudy.Results show this methodology is feasible.This methodology has adopted in China since 1998.And the experience shows that the stratification schema is efficiency,and the two individual sample frameworks can generate accurate estimation of crop acreage.

通过分析遥感技术在中国农作物种植面积估算中所遇到的难点 ,针对运行化的农作物遥感估产系统对主要农作物种植面积估算的需求 ,提出在农作物种植结构区划的基础上 ,采用整群抽样和样条采样技术相结合的方法 ,进行农作物种植面积估算。整群抽样技术利用遥感影像估算农作物总种植成数 ,样条采样是一种适合中国农作物种植结构特征的采样技术 ,用于调查不同农作物类别在所有播种作物中的分类成数。在中国现有的耕地数据库基础上 ,根据两次抽样获得的成数 ,计算得到具体某一种农作物类别的种植面积。最后给出了 2 0 0 3年早稻种植面积估算的实例。

... 种植现象普遍存在^[29] ...