Apple tree extraction and spatial analysis based on Sentinel-2 Image——A case study of Pingliang,Gansu Province
Using remote sensing technology to quickly monitor orchards and accurately grasp the area and spatial distribution of apple orchards can help promote local economic development.At present,there is relatively little research on the extrac-tion of orchards in hilly areas,and the effectiveness and reliability of related methods were still controversial.Taking Pingli-ang City,Gansu Province as the research area,such indicators as NDVI,RVI,EVI,SIPI,LSWI,and NDWI were used to enhance the input data.The gradient boosting tree algorithm based on data augmentation was used to extract the orchard planting area in the research area.To verify the effectiveness of the method proposed in this article,four machine learning algorithms,namely the minimum distance method,CART decision tree method,support vector machine method,and ran-dom forest method,were introduced for comparative analysis.The classification results showed that the gradient boosting tree algorithm had the highest classification accuracy,with an overall classification accuracy(OA)of 89.3%and a Kappa coefficient of 0.77.The classification performance and consistency were the best.In addition,the gradient boosting tree method based on data augmentation was used to extract the changes in orchard planting in Pingliang City from 2019 to 2023.The planting area of orchards in each district and county shows an overall upward trend,except for Jingchuan County.Jing-chuan County and Jingning County have the largest planting area,followed by Zhuanglang County,Lingtai County,and Kongtong District,and the smallest are Chongxin County and Huating City.
Remote sensingGradient boosting treeData augmentationSentinel-2 remote sensing imageKappa coeffi-ciencePingliang City
NETL
NSTL
万方数据
