Abstract: The intercropping system of tree-cotton is widespread in Xinjiang because it may increase yield and revenue especially during the early years of tree plantations.The statistics of the intercropped area is a key element for yield estimation.A method which can extract the tree-cotton intercropped ratio from planting area themetic map is proposed in this paper.The VHR (very high resolution) QuickBird imagery and multispectral high spatial resolution (GF-1) data were combined for extracting the intercropped ratio using the object-oriented approach and multi-seasonal classification approach respectively.Farmland extraction is a critical step to produce the intercropped information.Since multi-resolution segmentation (MRS) has been proved to be one of the most successful image segmentation algorithms, the trial-and-error process has been used to determine the three main optimal segmentation parameters (scale, shape, compactness) to identify farmland and tree canopy hierarchically.The new rule sets which consider optimal,shape and semantic features have been implemented to compile the farmland thematic map.Then, the GLCM-based texture feature has been proposed to distinguish the tree canopy when the image is segmented again.Intercropping ratio in each crop segmentation unit is calculated by stacking the farmland themetic layer and the tree canopy layer together.Since then, multi-seasonal classification approach has been used to extract the tree-cotton intercropping ratio from the intercropping ratio map.In addition, this work presents two varying images composed of GF-1 and Landsat8.By analyzing the phenologycal calendar, optimum temporal periods for cotton and other major crops are initially determined.Cotton planting areas are extracted by field samples supported supervised classification.The GF-1 accuracy achieves 89.16% which is by far better than TM results.Finally, tree-cotton interplanting area and ratio are extracted based on tree-crop intercropping map and cotton planting map.