Abstract: ObjectiveBased on multi-temporal GF-1 image,this paper extracted the planting area of staple crops in Jiangzhou,Chongzuo,Guangxi by using object-oriented classification,and provided reference for crop information extraction in southern hilly region.MethodBased on GF-1 image with 2 m resolution,human-computer interaction was conducted to accurately identify parcels information under land coverage.Cloud shadow detection was conducted based on multi-temporal GF-1 image.Features of the generated image such as spectrum,normalized differential vegetation index (NDVI) and brightness were processed.Crop information of sugarcane,rice and banana were extracted using object-oriented classification method.ResultAccording to the result of confusion matrix evaluation classification,the overall accuracy of staple crops extraction was 90.08%,and Kappa coefficient was 0.85,which met the accuracy requirements for application of agricultural achievements.ConclusionThis paper uses the effective image data and combines parcel data to complete the extraction of crop information.This technical method can accurately extract crop information in hilly regions,and lay a foundation for solving the problem of extracting crop information in rainy hilly region in the south.