Study of Object-oriented Extraction of Main Features along Reconstructed and Expanded Highway with Airborne DOM
Aiming at the problems of great difficulty in extracting feature information along the reconstruction and expansion highway and the tight schedule,we take the airborne high-resolution DOM remote sensing image of the Chengdu south expressway expansion project(near Chengdu Luoshiba interchange)as the research object,and use three object-oriented classification methods,such as KNN,CART,SVM,etc.,to extract and classify the main feature information in the belt area within the left and right of this highway within the range of about 300m.The results show that:1)the determination of the optimal segmentation scale is a complex process and is constrained by many factors.In this paper,the ESP(Estimation of Scale Parameter)tool is used to simplify the acquisition of the optimal segmentation scale,which achieves good image segmentation results;2)the three object-oriented classification results basically meet the accuracy requirements for the pre-planning,design,and construction of the reconstruction and expansion of highway projects;the three methods are based on the same training samples and validation samples.In terms of classification accuracy,SVM>CART>KNN,and in terms of specific features,the classification accuracy of features such as residential land,transport land,vegetation and water bodies is relatively high,and the classification accuracy of features such as bare land,sheds and other features is relatively low;3)the three classification methods are not only technologically advanced and economically efficient,but also have significant practical value for the automated and intelligent extraction of features along the reconstructed and expanded highway and their application in digital transport.
airborne DOM imagereconstruction and expansion of expresswayobject-oriented classificationfeature extraction
NETL
NSTL
万方数据
