Skip to Main Content
Impervious surface is an important factor for evaluating the process of urbanization and its rate. In urban areas, there are a large number of street trees obscuring impervious surface, which will cause underestimation of impervious surface if it is directly extracted by classification of remotely sensed imagery. In order to solve this problem, this paper presented an approach to reduce the impact caused by street trees. The method combined very high resolution (VHR) image and urban street map. Experiments were conducted using a VHR multispectral QuickBird image acquired in Nanjing, Jiangsu Province, China. First, a multiscale and multilevel segmentation method was used to conduct object-based image classification , separately using QuickBird image and shadow areas in the image. Second, the buffer of street region with a certain width was created, which was then overlaid with the classification result to analyze and extract impervious surface from tree class. The proposed method reduced impact of street trees to impervious surface, all the land cover classes were finally combined to two classes: pervious surface class and impervious surface class. Finally two methods were used to evaluate the results of impervious surface extraction. First, the classical confusion matrix for classification accuracy was used. Second, calculation and analysis of the impact was conducted. Results from the experiments indicated that the proposed method obtained more accurate results than that from direct classification. After adopting the proposed method, the area of impervious surface was increased by 20.49% for the study area.
Date of Conference: 16-18 Dec. 2012