Study On Classification Method Of Main Forest Types In Badaling Forest Farm Based On GF-2

The successful launch of the GF-2 satellite indicates that the Chinese remote sensing satellites enter the sub-scale high-yielding era.Based on the images of GF-2,this paper studies the processing of GF-2 image,the classification of forests in northern collective forest regions by means of remote sensing.Through analysis and research,conclusions of this research paper were as follows(1)In the process of forest type classification,we chose different scale and merge experiments,and select the appropriate segmentation and merging threshold for different extraction objects.We chose the segmentation and merging threshold of 40 for segmentation based on edge detection.The multi-scale segmentation based on region was chosen as the threshold of segmentation and merging,which improved the accuracy of forest type classification(2)In the study of forest type classification,the object-oriented forest type classification method was used,and the accuracy of different segmentation methods under the same classification algorithm were compared.In the case of the same sample,the overall accuracy of the segmentation method based on edge detection was 62.43%,The kappa coefficient was 51.99%,the overall accuracy of the multi-scale segmentation based on the region was 71.68%,and the kappa coefficient was 64.54%.(3)In the case of the same sample and the same segmentation method,the classification accuracy of the different classification methods.The overall accuracy of the random forest method was 75.43%,the kappa coefficient was 68.63%,and the SVM was classified according to the multi-scale segmentation method.The overall accuracy of the method was 66.76%and the kappa coefficient was 58.75%.(4)The results showed that there were quite a lot confusion between shrub and broad-leaved forests,as well as the pinus and broad-leaved forests shadows.