基于HJ-1A/1B CCD时间序列影像的水稻生育期监测

摘要

为了准确获得作物大面积生育期信息,指导农业生产管理和决策,基于2009、2010年湖南双季稻生育期间42景晴空HJ-1A/1B CCD影像,提出了一种基于小波分析的生育期估计方法:首先通过线性插值构建像元尺度上逐日的增强植被指数时间序列集,随后采用小波分析对其进行去噪重建,得到反映水稻生长过程的日时间序列曲线,最后通过曲线特征估计水稻的关键生育期.以抽穗期为例,利用湖南沅江市20个田块实地调查数据对估计结果进行了验证,并与基于非对称高斯函数及基于Savitzky-Golay滤波的方法进行了对比,结果表明基于小波分析的方法平均误差为3 d左右,整体上不高于后2种方法误差,能较精确地估计双季稻抽穗期,提供了一种新的有效监测水稻生育期信息的方法.%Crop growth stages information in large-scale is crucial to guide the management and decision-making in agriculture production.In order to obtain a precise information, a wavelet-based method for estimating rice growth stages was proposed based on 42 cloudless HJ-1A/1B CCD images in Hunan's double cropping rice growth period in 2009 and 2010.Firstly, a time-series EVI dataset at day frequency at pixel scale was built by linear interpolation.Then a profile, which reflects the rice growth process, was produced by filtering based on wavelet analysis.At last, the critical growth stages could be estimated according to the characteristics of the curve.As an example, 20 parcels in situ investigation for rice heading date in Yuanjiang City were used to verify the results, and the wavelet-based method was compared to those based on non-symmetrical Gaussian function and Savitzky-Golay filtering.The result showed that the proposed method, with an average error of 3 days which was less than the latter’s errors, could accurately monitor heading date of double cropping rice.This method is available in monitoring rice crop growth stages.