Application of Gaofen-6 WFV in complex mountain feature classification
Aiming to evaluate its effectiveness in multi-class land cover classification,the GF-6 WFV imagery was used to conduct large-scale land cover classification research in the complex mountainous region of southwest Sichuan.By combining spectral bands and calculating vegetation indices,the ability to monitor vegetation health was enhanced.The introduction of the red-edge band(B5)and yellow band(B8)provided new technical advantages for vegetation and land use classification.In terms of supervised classification methods,four approaches were employed,including Mahala-nobis distance,Maximum likelihood,Convolutional neural networks(CNN),and Support vector machine(SVM).The re-search indicated that SVM performed exceptionally well in handling high-dimensional spectral data and complex terrain conditions,achieving the highest classification accuracy.The classification accuracy of the Mahalanobis distance and Maximum likelihood methods was lower,mainly due to data assumptions and sample size limitations.The performance of the Neural network method was suboptimal,primarily due to insufficient quantity and diversity of training samples,which resulted in weak model generalization.The results demonstrated that GF-6 WFV imagery exhibits superior perfor-mance in land cover classification,particularly in precision agriculture and forestry management.Future research should focus on integrating multi-source remote sensing data,optimizing algorithms to improve classification accuracy,and re-ducing computational resource consumption.
mountain feature classificationwide field of viewmultispectral informationred-edgeSupport vector machine classification
万方数据
维普
