Improving Urban Vegetation Classification Accuracy With Multispectral Imagery and Lidar
Date
2013-08
Authors
McDaid, Guinevere
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Urban areas are comprised of fine-scale heterogeneous land-cover classes and detailed land cover classifications often require multiple techniques and classification methods to produce an accurate land cover land-use map. Policy makers and urban developers need up-to-date, precise data in which to base decisions and to guide development decisions that meet multiple objectives. Accurate and up-to-date land cover data, particularly in rapidly developing cities, is often unattainable at the spatial resolution desired by urban planners. Aerial remote sensing is a suitable and effective source for urban land cover mapping as the image datasets for classification are acquired at a high spatial resolution (e.g., 1 m). This study examines added utility of integrating1 m image data, lidar height data, and lidar intensity data as a means of increasing the classification accuracy of urban vegetation classes compared with that of a classification using aerial image data alone. One meter National Agricultural Inventory Program (NAIP) data, acquired in 2010 were used as input to an object-oriented, supervised classification in ENVI EX to derive urban vegetation land cover in the downtown area of San Antonio, Texas. Classification of data adhered to the Texas Land Classification System (TXLCS). The classes used here include developed, developed open-space, broad-leafed evergreen, cold deciduous, mixed forest, and shadows. These analyses indicate that the addition of lidar height data as a classification layer did not improve classification accuracy compared to image data alone. The addition of lidar intensity data as a classification layer did however improve the classification accuracy compared to image data alone. The integration of spectral and intensity data does produce a more accurate urban vegetation land cover map.
Description
Keywords
Urban vegetation, Multispectral imagery, Lidar, Classification accuracy, Object oriented
Citation
McDaid, G. (2013). <i>Improving urban vegetation classification accuracy with multispectral imagery and lidar</i> (Unpublished thesis). Texas State University-San Marcos, San Marcos, Texas.