Urban Land Use Classification from UAV Remote Sensing Images Based on Digital Surface Model
SONG Xiaoyang;HUANG Yaohuan;DONG Donglin;ZHANG Fei;College of Geoscience and Surveying Engineering, China University of Mining & Technology;State Key Laboratory of Acoustics, Institute of Acoustics, Chinese Academy of Sciences;State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences;University of Chinese Academy of Sciences;
Urban land use is a key issue of urban ecology. It is of great significance to understand the urban land use for planning urban functional zones, improving land use efficiency, estimating human population, analyzing urban landscape and promoting regional economic and environmental development. Therefore, urban land use classification research has been one of the core contents of urban planning and urban geography. With the rapid development of Unmanned Aerial Vechicle(UAV) technology, rich UAV data have widely been used in different kinds of fields, especially in the urban land use classification. Digital surface model(DSM) and digital orthophoto map(DOM) obtained from UAV remote sensing images can effectively improve the accuracy of urban land use classification. In order to make full use of the rich information of UAV remote sensing images, an urban land use classification method is proposed using high-resolution DOM and DSM. In this study, the composite bands of DOM and DSM were used as data source. Considering the characteristics of urban land use, the object-oriented classification method was optimized by combining DOM spectral information with DSM, which is used as the final threshold of the pixel merge in multi-resolution segmentation and as height feature in objects classification,respectively. The method was validated in Jingjinxincheng located in Baodi District, Tianjin City. The results showed that, comparing with the initial multi-resolution segmentation method, all of the segmentation quality rate(QR), over-segmentation rate(OR), under-segmentation rate(UR) and comprehensive rate(CR) of optimized multi-resolution segmentation method were reduced, and the effects of image segmentation has been improved significantly. The optimized object-oriented classification method improved the classification accuracy,especially for the extraction of roads, buildings and other constructions. The overall accuracy of the classification results increased from 85% to 87.25% and the Kappa coefficient also increased from 0.79 to 0.82. Therefore, the optimized object-oriented classification method can be used for urban land use study more effectively.
【CateGory Index】： P237
【CateGory Index】： P237