Time series of NDVI data have extensive applications in crop condition monitoring and yield prediction.The NDVI data are polluted by cloud even with precise correction and 10 days composition. And due to its 1 km resolution

the data of one pixel contains information more than crops themselves. In this paper

we present a method to extract crop growth profile using time series of SPOT-VGT NDVI and land cover data. In order to remove cloud pollution and extract the characteristics of the vegetation dynamics

the Harmonic Analysis of Time Series （HANTS） algorithm is performed on the time series of NDVI data.HANTS can analyze the time signal for each individual pixel and use the series of harmonic sine and cosine waves to reconstructe waves fitting into the total period.Results show that the cloud polluted data are recognized and replaced successfully by reconstruction of seasonal NDVI profiles. In order to monitor the crop condition and indicate yield level

the time series of NDVI should contain as much crop information as possible

from which the extraced indicators should be related to crop yield performances. Land cover data is used to mask non-arable land

and the NDVI are calculated only for arable land pixel.For a regional

like a county or a province

the pixels within a region are used only for those that NDVI is higher than 0.15. The weight average method is used to calculate the regional NDVI which can indicate the crop information

better than other methods. The weight is the proportion of arable land within a pixel. By comparing the profiles of whole country

a region

a province and a county

we can conclude that the profile is affected by the crop calendar and the cultivation structure. If a region has same crop calendar and pure crop structure

the crop profile is a good indicator for crop condition and yield. The yield stratification zone and county level are proposed to calculate regional NDVI profile. The profiles are validated by using field measuring LAI.It shows that HANS can increase the relative coefficient of NDVI and LAI by 5%—11% than raw data. The weight average method can increase the relative coefficient of NDVI and LAI by 14%—17% than other methods.