Abstract: Crop classification via remote sensing is valuable to obtain spatial distribution of crop types, capture crop sowing information, which is important for managing agricultural production and making agricultural policies. Remote sensing is advantageous over conventional methods to investigate crops in a large scale. However, some problems also need to be worked out, because different objects have a similar spectral response and same objects with different spectra. To avoid this kind of problem in one single temporary image, to improve the accuracy of crop classification result via MODIS images, and to show more details in comparison of classification result by MODIS imagery, we used MODIS NDVI time series data to select suitable date for distinguishing crop types, based on multi-temporal Landsat8 OLI imagery, which have higher radiation resolution to extract crops information, and applied object-oriented classification method with spectral and phenological information to classify crop types. In addition, we applied this method in Bei'an county, an important area for grain producing region in Heilongjiang province. After all, we got spatial distribution map for main crops in Bei'an county. Then we selected 377 ground-truth points that represent all kinds of ground objects, to evaluate the precision of the classified results. It showed that, this method considerably increased the classification accuracy. Consequently, the total precision was 90.7%, and the kappa index was 0.88. It can be concluded that the method of combining Landsat8 OLI imagery with MODIS NDVI time series, based on object-oriented classification method can improve accuracy and increase efficiency of crop classification, which has enormous potential. It can be applied to other regions, or to map distribution of other ground objects.