Yaogan jishu yu yingyong · 2011 · 14 citations · 0 references
Natural SciencesEarth SciencesGlobal Navigation Satellite SystemsRemote SensingTime Series AnalysisLand Surface MonitoringMulti-temporal Image AnalysisTime Series ModelingVegetation Change DetectionVegetation IndicesVegetation MonitoringMean Value AnalysisSatellite Image Time Series AnalysisSatellite Vegetation Index AnalysisSpatial-temporal Vegetation AnalysisTemporal Vegetation Change DetectionTime Series Remote Sensing AnalysisTime-series Vegetation Monitoring
MODIS 16 days composited NDVI time-series of 2007-2009 in northern Tibet are taken as study case to compare the characters of three principal fitting methods,that is the double Logistic function fitting(D-L),asymmetric Guassian function fitting(AG),and Savitzky-Golay filtering(S-G) methods.To begin first,the basic principles and implementation process of the three algorithms are introducted based the TIMESAT 2.3 program.Then,the results of NDVI time-series fitting based on the three methods are emphatically compared and analyzed,from the aspects of fitting effect for the upper envelope curve of original NDVI series and the ability of preserving the high-quality NDVI fidelity.The results show that the three fitting methods would raise the mean value of NDVI samples to some degree,and that the AG and D-L fitting algorithms generate more consistently reconstructed NDVI time-series to the original NDVI temporal curve than the S-G filtering method.Secondly,the fitting NDVI values of AG and D-L methods are higher than points of the upper envelope curve,the S-G filtering method is opposite.Among the three algorithms,AG fitting produce the most approximative results.Besides,the AG and D-L methods perform extremely similarity to keep the fidelity of high-quality NDVI samples,and their fitting NDVI series are better than that of S-G filtering except the peak period of growing season.