The information extraction of Gannan citrus orchard based on the GF-1 remote sensing image

International Symposium on Earth Observation for One Belt and One Road

地球科学;政治学;社会学;经济学

Wang, S.^1 ; Chen, Y.L.^1

School of Architectural and Surveying Engineering, Jiangxi University of Science and Technology, Ganzhou

341000, China^1

关键词: Classification accuracy;    Classification methods;    Classification results;    Decision tree classification;    Hierarchical decision trees;    Industrial planning;    Multiscale segmentation;    Remote sensing images;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/57/1/012001/pdf
DOI  :  10.1088/1755-1315/57/1/012001

来源: IOP

【 摘 要 】

The production of Gannan oranges is the largest in China, which occupied an important part in the world. The extraction of citrus orchard quickly and effectively has important significance for fruit pathogen defense, fruit production and industrial planning. The traditional spectra extraction method of citrus orchard based on pixel has a lower classification accuracy, difficult to avoid the "pepper phenomenon". In the influence of noise, the phenomenon that different spectrums of objects have the same spectrum is graveness. Taking Xunwu County citrus fruit planting area of Ganzhou as the research object, aiming at the disadvantage of the lower accuracy of the traditional method based on image element classification method, a decision tree classification method based on object-oriented rule set is proposed. Firstly, multi-scale segmentation is performed on the GF-1 remote sensing image data of the study area. Subsequently the sample objects are selected for statistical analysis of spectral features and geometric features. Finally, combined with the concept of decision tree classification, a variety of empirical values of single band threshold, NDVI, band combination and object geometry characteristics are used hierarchically to execute the information extraction of the research area, and multi-scale segmentation and hierarchical decision tree classification is implemented. The classification results are verified with the confusion matrix, and the overall Kappa index is 87.91%.