Multi-temporal vegetation index reflects the phonological calendar features which are one of the most distinguishing features between different crops.The main weakness of which is the difficulty of comprehensive utilization of multi-source remote sensing data in automatic crop mapping.In this paper,Xinjiang Bole was taken as studying area; historical vegetation index profiles of different kinds of crops were built from MODIS data set; a medium to high spatial resolution time series vegetation data was built from TM and HJ-1 dataset after spatial scaling based on MODIS data set and the Euclidean distance was calculated between the reference time series profile and the profile of the medium to high vegetation index data set to map the crop pattern automatically.The result show that the accuracy of mapping a high spatial resolution time series data built from multi-source remote sensing data set based on reference profiles is above 75％,which is much higher than the accuracy of crop pattern mapping only using MODIS data set.