摘要
鉴于作物类型识别中存在光谱特征相似的困扰,"异物同谱"问题难以有效解决,而时序归一化植被指数(Normalized Different Vegetation Index,NDVI)曲线数据能够反映作物不同时期的动态变化趋势,该文将NDVI时间序列投影到N维空间构成多维特征矢量,结合冬小麦特有的物候特征,充分利用矢量的方向和大小参量,构建冬小麦识别的矢量分析模型,模型的识别能力较强,可以充分发挥NDVI时间序列的优势。以唐山市为研究区,基于高分一号WFV(Wide Field of View)数据的高分辨率优势,构建覆盖冬小麦生长期的NDVI时间序列,采用矢量分析模型进行冬小麦识别,同时与最大似然法、马氏距离法、支持向量机法、神经网络法、最小距离法等分类方法进行对比。结果显示,后5种分类方法的Kappa系数介于0.701 8和0.790 3之间,而矢量分析模型达到了0.895 2,精度有了较大提高。该研究为冬小麦识别提取提供了新的思路,也对推动遥感农情信息调研具有一定学术和应用价值。同时,基于研究区训练样本提出了模型阈值参数自动确定的方法,为今后冬小麦自动提取奠定了基础。
The puzzle of similar spectral characteristics exists in crop identification,and problem of foreign bodies with spectrum cannot be effectively solved.As time series of NDVI curve can reflect the dynamic change trends of crop in different periods,this paper uses the vector analysis method to map NDVI time series into N dimensional data space to constitute the different characteristic vector.On the basis of the mathematical theory of vector,combined with winter wheat phenological characteristics,construct vector analysis model of winter wheat information identification by making full use of direction and size of vector.Model with strong identification ability can give full play to the advantages of NDVI time series.The paper takes Tangshan city as the study area,uses the high resolution advantages of GF-1/WFV（Wide Field of View）multi-spectral data,constructs NDVI time series which completely covers the growth period of winter wheat,and then adopts the proposed model to identify the winter wheat information.By comparing the experimental results of maximum likelihood method,Mahalanobis distance method,support vector machine method,neural network,and minimum distance method,classification results show that the Kappa coefficient of other methods is between 0.701 8and 0.790 3,while Kappa coefficient of vector analysis model reaches0.895 2,and the proposed model greatly improves the precision of winter wheat identification.This research can provide new ideas and methods for the extraction of winter wheat information,and it also has important significance on science and worthiness in practical application for agricultural information monitoring.Meanwhile,this paper presented a method based on the training samples in the study area to automatically determine the division threshold of the proposed model,which can lay the groundwork for future winter wheat automatic extraction.
出处
《遥感信息》
CSCD
北大核心
2016年第5期53-59,共7页
Remote Sensing Information
基金
国家自然科学基金(41371416)
国家科技重大专项资助项目(03-Y20A04-9001-15/16
30-Y20A29-9003-15/17)
关键词
冬小麦
NDVI
时间序列
矢量分析模型
自动识别
winter wheat
NDVI
time series
vector analysis model
automatic identification
