Reconstruction Of NDVI Data With High-spatial And Temporal Resolution Using Spatio-temporal Fusion Technique

The normalized difference vegetation index (Normalized DifferenceVegetation Index: NDVI) is an important indicator of vegetation growth status,seasonal changes and the vegetation coverage, and can directly reflect thegrowth process of vegetation changing with the time, so it has been widelyapplied in research on land cover change, vegetation and phenology at global orregional level. Although the NDVI data with both high spatial resolution andhigh time resolution (high time series) can be more accurate to describe thetemporal variation characteristics of vegetation and other objects after datareconstruction, but in view of noise factors such as the influence of atmospheric,cloud pollution and sensor limitations of the imaging system, it is difficult todirectly obtain NDVI data with high spatial and temporal resolution. Thus, in theexisting technical conditions, developing a kind of technological strategy thatcan be used for production and analysis of NDVI data with high spatial andtemporal resolution is the key scientific problem in the application above.As a new research hotspot in the multi-source remote sensing data fusion field, temporal spatial fusion technology can effectively integrate the spectralchange information in space dimension and time dimension, thereby to get thesynthetic multispectral data with high spatial and high temporal resolution byusing remote sensing data with low spatial and high time resolution. This paperputs the temporal spatial fusion technology research as the starting point, and isto seek the optimum strategies in the process of analysis application of NDVIdata through analyzing the prediction precision of generation and reconstructionalgorithm of NDVI data with high spatial and temporal resolution for typicalspectral features. According to the research goal, the research content of thispaper includes the following three aspects:1) In view of the characteristics of regional application of NDVI data andthe limitation of current application based on temporal multi data fusion model,this paper takes the space-time fusion strategy based on single data as technicalbasis, according to this, make experimental analysis and comparison on thecurrent two kinds of main spatial-temporal fusion model based on reflectance(adaptive fusion model and semi physical fusion model), although the predictionaccuracy for the specific band of the semi physical fusion model is higher, butthe prediction t of the STARFM algorithm for NDVI is better;2) The three existing common reconstruction method of time series data(asymmetry Gauss function fitting, double Logistic function fitting and the S-Gfiltering method) will be deeply analyzed in the experiment. This paper willcomprehensively compare the reconstruction precision of the above methods from qualitative and quantitative angle, all the three kinds of data reconstructionmethods can remove most of the noise and make the original curve more smooth,the fidelity of S-G filter is the best;3) Take Landsat TM and MODIS reflectance data in a particularexperimental area as an example, through analyzing the different processescombination of the above two kind of temporal spatial fusion model and threekinds of data reconstruction algorithm and the prediction effect of eachprocess combination for the timing feature of NDVI data in the the study area,can initially summarize The availability of NDVI time series data and its timesequence features are obtained by STARFM algorithm and S-G datareconstruction method.