Research on the Extraction Method of Apple Orchard Information in Florescent Based on GF-1 Image-A Case Study in Yiyuan County

Data are increasingly available from new multi-spectral and high-resolution remote sensors for extracting agricultural information. We conducted a study of extracting apple orchard information from the GF-1 images. Using Yiyuan County in Shandong Province, China as the study area, we compared several different classification methods, including an integrated unsupervised-supervised classification method (improved maximum likelihood), the support vector machine (SVM) method, the nearest-neighbor method, and the decision-tree method. Particularly, we applied the first two methods to the pixel-based classification, and applied the last three methods to object-oriented classification. The comparison finds that the last three methods have higher efficiencies, whereas the first two methods have higher classification accuracy (according to the confusion matrix).