Study On The Information Extraction Method Of Citrus Orchard Based On Object-oriented

China is one of the main producers and consumers of citrus fruits,which has a history of more than 4000 years of cultivation.Extracting of citrus fruit fast and effectively has important significance for fruit pathogen defense,fruit production and industrial planning.The traditional extraction method is based on pure spectra of pixel,which classification accuracy is very low and it is hard to avoid “salt and pepper”.The phenomenon of different objects with the same spectrum is more serious affected by noise.Containing more texture information,shape information and context information is prominent features of high resolution remote sensing images.Taking Xunwu County of Ganzhou city citrus fruit planting area as the research object,aiming at the shortcomings of traditional classification method based on pixel,do the following research based on the GF-1:1.In order to make the remote sensing image data of GF-1 not only containing multi-spectral characteristics but improving spatial resolution,do some image preprocessings which are atmospheric correction,ortho correction and image fusion on the data.2.On the premise of multi-scale segmentation,various feature information of objects have same certain change rules.The change of the mean value,standard deviation and the NDVI value of vegetation objects based on the scale of segmentation.3.Choose the best area ratio method to measure the pros and cons of a particular object segmentation object level.4.Study on neighbor classification model in-depth based on fuzzy mathematics method,based on mathematics expression in-depth analysis model,full of statistics analysis on whole site image of spectrum features.5.It constructed an object-oriented decision tree classification model combined with fuzzy mathematics theory.Based on the multi-scale segmentation establishment of object-level organizational structure,the nearest neighbor classifier was integrated,and the hierarchical classification model of object-oriented decision tree was established by fusion fuzzy mathematics method and decision tree model.Firstly,the spectral features and texture features are selected by statistical analysis.Then,use the single band threshold,NDVI,band combination,object geometry and texture features to analyze the information to realize multi-scale segmentation and hierarchical decision tree classification.The classification results were verified using the confusion matrix,the Kappa coefficient is 0.8643,especially significantly improve the extraction accuracy of fruit trees.