夏玉米生长过程曲线重建研究——以鹤壁市为例

增强型植被指数(EVI)时间序列数据(即植被生长曲线)是整个生育期内植被各种生物学特征的综合反映。由于太阳位置、大气、地表和传感器位置与性能等的影响,根据遥感数据计算的EVI值往往比实际值偏低(存在大量噪声),并不能反映植被生长的真实情况,应用前需进行去噪重建工作。针对目前生长曲线重建研究大多是针对MODIS等国外遥感数据的情况,在综合分析重建方法的基础上,利用风云3号卫星的MERSI中分辨率遥感卫星数据构建鹤壁市夏玉米的EVI生长过程曲线。首先,用最大值合成法(MVC)对原始EVI时间序列数据进行初步的去云处理。接着,利用基于时间域的Savitzky-Golay滤波(简称SG滤波)对该EVI序列进行进一步的平滑去噪处理,结果发现,在噪声点EVI数值提高了,但同时在其他不是噪声点的地方EVI的值降低了。针对这种不合理的情况,利用基于SG迭代滤波取上包络线的改进方法进行处理,很好地克服了上述缺陷,在非噪声点EVI数值适当提高,且曲线平滑,达到了生长曲线重建的目的。然后,采用基于频率域的小波变换方法进行实验对比,结果发现,小波变换存在着与经典SG滤波类似的缺陷,而且在曲线末端存在突变情况。经过比较分析发现,针对研究区的实际情况,改进SG迭代滤波是较优的去噪方法。

Research on Summer Corn Growth Curve Reconstruction: A Case Study in Hebi

Enhanced Vegetation Index (EVI) time series data is the comprehensive reflection of various vegetation biological characteristics in the whole growth period. Since the influences of some factors such as solar position, atmosphere, ground surface, and the position and property of sensors, the EVI value based on remote sensing data is usually lower than the actual value because of a large number of noises. Therefore it is not able to reflect the real growth situation of vegetation. The de-noising reconstruction work should be done before it is applied. Given the situation that the growth curve reconstruction researches at present mostly using foreign remote sensing data such as MODIS. On the basis of comprehensive analysis of reconstruction methods, the Medium Resolution Spectral Imager (MERSI) data from FY3 satellite was used to construct the EVI growth curve of summer corn in Hebi. Firstly, preliminary cloud removing for original EVI time series data was done by Maximum Value Composites (MVC). Then, time domain based Savitzky-Golay filter (SG filter) was used to do further noise smoothing for the EVI series. The results show that the EVI values increase at noisy points, meanwhile decrease at non-noisy points, which could not completely agree with the fact. In order to solve the problem, the envelope curve of the EVI series was constructed by utilizing SG iterative filter as an improved method. Then the EVI values could properly increase at non-noisy points, and the curve could be smooth. So that the purpose of summer corn growth curve reconstruction was achieved. In addition, experimental comparison was done between frequency domain based wavelet transform method and the SG iterative filter method. The results show that the wavelet transform method has defects which are similar with classical SG filter, and mutation exists at the end of the curve. Through comparative analysis, it shows that according to the fact in study area, the revised SG iterative filter method is the relatively optimal de-noising method.