An Integrated Approach to Agricultural Crop Classification Using SPOT5 HRV Images

An integrated method that incorporates the advantages of per-parcel and perpixel approaches as well as spectral and spatial characteristics was proposed for crop classification of a typical agricultural area in south-east China using SPOT5 HRV data. The co-occurrence texture was employed to evaluate the heterogeneity of the image data. The average parcel textures determined each parcel defined by the crop boundaries to be classified whether on a per-parcel or per-pixel basis. The optimal threshold in the span of texture ranges was detected by trend analysis, which assigned the proportions of each approach in the integration, thus to produce the best integrated classification. It was suggested that this integrated approach can be effectively implemented to produce crop classification maps with higher accuracy from satellite images of medium and high spatial resolution in a complex agricultural environment, where both homogeneous and heterogeneous crop fields occur side by side.