International Journal of Applied Earth Observation and Geoinformation
Extraction of multilayer vegetation coverage using airborne LiDAR discrete points with intensity information in urban areas: A case study in Nanjing City, China
Introduction
Urbanization has been an important social and economic phenomenon and it is taking place at an unprecedented scale and rate in the developing countries, for example China (Sun et al., 2013). Urban vegetation ecosystem provides important environmental services including air and water purification, wind and noise filtering and microclimate stabilization (Chiesura, 2004). In addition, urban ecosystem provided social and psychological services for the livability of cities and the well beings of the urban inhabitants.
LiDAR (Light Detection and Ranging) has recently become an effective means of acquiring 3D information (Priestnall et al., 2000, Stephen and Yong, 2003, Vosselman et al., 2005, Wang and Glenn, 2009, Hartfield et al., 2011, Ramdani, 2013). Extraction of vegetation coverage is an important part of LiDAR application in urban areas. It is possible to obtain the type of wetland vegetation and its distribution using LiDAR data (MacKinnon, 2001). Many researchers have focused on LiDAR data combined with the other high-resolution optical images (Ramdani, 2013), for example Quickbird multi-spectral data, CALMIT AISA hyperspectral data and LiDAR data to identify urban trees species (Sugumaran and Voss, 2007). LiDAR and Quickbird data have been used to evaluate urban tree shadows in residential and commercial districts (Tooke and Voogt, 2009). Land use and land cover classification have been improved in urban areas by integrating remotely sensed multispectral reflectance data and LiDAR-derived height information (Hartfield et al., 2011). Urban buildings can be detected out of vegetation category through the combined use of LiDAR data and Quickbird imagery (Vosselman et al., 2005, Chen et al., 2012). However, the available studies mainly focused on horizontal vegetation coverage and used LiDAR data with optical images.
More recently, the 3D detection of vegetation coverage in urban areas has become possible by using airborne LiDAR data. LiDAR sensors currently used for 3D vegetation mapping and forestry research can be categorized as either discrete return or full waveform systems (Lim et al., 2003, Parrish and Scarpace, 2007). A 3D analysis of LiDAR waveforms has been developed to characterize the total volume and spatial organization of vegetation and empty space within a forest canopy (Lefsky et al., 1999). Iovan and Cord (2007) put forward an automatic acquisition method of urban vegetation information that combined high-resolution aerial images and DSM by LiDAR data. Hecht et al. (2008) calculated the volume of urban vegetation and analyzed the data accuracy using single echo LiDAR data and fuzzy theory. Using images and the point cloud, Höfle and Hollaus (2010) has extracted urban trees from high-density airborne full-waveform LiDAR data using an object-oriented method. Wagner et al. (2008) has classified a vegetation canopy and the underlying terrain using echo information of airborne full-waveform LiDAR data in an urban area. Advanced discrete return sensors can represent complex structures of vegetation targets at a level of detail equivalent in some aspects to the content of full-waveform data (Ussyshkin and Theriault, 2011). Jensen et al. (2008) have predicted the leaf area index based on discrete return LiDAR in two conifer forests.
LiDAR intensity may be used to separate green vegetation from fields and asphalt (Maas, 2001). Three main factors influence the backscatter of the emitted laser power: (a) spherical loss, (b) topographic effects and (c) atmospheric effects. These factors lead to a noticeably heterogeneous representation of the received power (Höfle and Pfeifer, 2007). Various methods have been proposed to normalize the intensity. Langford et al. (2006) explored small footprint LiDAR intensity data in a forest. Kaasalainen et al. (2009) used commercially available sand and gravel as reference targets and calibrated these reference targets under laboratory conditions to determine their backscatter. The intensity values are marginal according the emitted laser power in the test laser data. However, little emphasis has been placed on how the intensity of LiDAR data can be used as an information source for vegetation applications (Wagner et al., 2008).
Up to now, most of the laser data were acquired by discrete return instead of full waveform commercial LiDAR systems, especially in developing countries, such as China. The enormous amount of data need processed efficiently and accurately. In this paper, a novel workflow used laser intensity to classify urban vegetation was put forward, and it was applied in Nanjing City, China. The results were compared with: (1) classification using Quickbird images qualitatively and (2) classification using TerraSolid software (Terrasolid Ltd., Helsinki) methods quantificationally. Then, two layers vegetation coverage, trees and shrub (including grass) were accessed using the inverse distance weighting (IDW) interpolation method with classified laser points.
Section snippets
Study area and data
The area of this study is located in the center of Nanjing City (31°14′–32°36′ N, 118°22′–119°14′ E), Jiangsu Province, Eastern China. Nanjing is one of representative ecological cities in China.
The red box shown in Fig. 1(a) was specifically selected. It is a 250 m by 250 m2 in the central part of the city, with a small park in the center surrounded by roads and high buildings. The area is characterized by a great variety of plant and tree species, including high trees, short-cut trees,
Methodology
The testing LiDAR data using in the paper were stored in the LAS (the ASPRS LiDAR data exchange format) files. The data were processed following the steps outlined in Fig. 2. Before using the intensity information of the laser points, high-frequency noise was eliminated using the median filtering method based on discrete points. The principle of the intensity values of scattered objects was observed and analyzed in the same airborne LiDAR mission. The point clouds were classified into
Extraction of multilayer vegetation coverage
The extraction of multilayer vegetation coverage requires two steps: (1) layering the laser points and (2) the interpolation of layered laser points.
The result of vegetation classification based on intensity information
The result of vegetation classification based on intensity information of laser point was shown in Fig. 5.
The red polygons labeled as 1 and 3 represent parts of roads in Fig. 5(a). Some non-vegetation points in the polygons are falsely classified as vegetation because the intensities of the painted lines on roads, vehicles or signal poles are greater than those of the vegetation. The red polygon labeled 2 represents a pavilion with several large layered cornices. The cornices of the pavilion
Discussion
In this study, the method is developed to classify vegetation and non-vegetation with intensity information of laser points. It is a first try to use LiDAR discrete points to extract multilayer vegetation coverage in urban environment in China where urbanization is happening at a scorching pace. This method has applied successfully in Nanjing city, eastern China to extract multilayer vegetation coverage. Urbanization happened very fast in China during the last decades. It to some extent
Conclusions
In this paper, multilayer vegetation coverage is extracted in our methodology by intensity information of airborne LiDAR discrete points in Nanjing urban area, China. Through validation of field investigation, it is shown that the extracted results have high performance of vegetation classification. Compared with results using the Quickbird image, there are few commission errors in large vegetation area by intensity information of airborne LiDAR discrete points. The LiDAR discrete points are
Acknowledgments
This research work was supported by the 973 Program (2010CB951503), 863 Program (2008AA12Z106), the Natural Science Foundation of China (40501047), and the Priority Academic Program Development of Jiangsu Higher Education Institutions. The authors thank Dr. Guorong Hu of Geoscience Australia, and Dr. Hong Yang of the School of Geography, University of Southampton, UK, for giving valuable suggestions on modifying this paper. The authors also thank the anonymous reviewers for providing valuable
References (32)
The role of urban parks for the sustainable city
Landscape and Urban Planning
(2004)- et al.
Correction of laser scanning intensity data and model-driven approaches
ISPRS Journal of Photogrammetry and Remote Sensing
(2007) - et al.
Testing LiDAR models of fractional cover across multiple forest ecozones
Remote Sensing of Environment
(2009) - et al.
Discrete return LiDAR-based prediction of leaf area index in two conifer forests
Remote Sensing of Environment
(2008) - et al.
A LiDAR-derived canopy density model for tree stem and crown mapping in Australian forests
Remote Sensing of Environment
(2007) - et al.
LiDAR remote sensing of the canopy structure and biophysical properties of Douglas-fir western hemlock forests
Remote Sensing of Environment
(1999) Adaptive vector median filtering
Pattern Recognition Letters
(2003)- et al.
Extracting urban features from LiDAR digital surface models
Computers, Environment and Urban Systems
(2000) - et al.
Quantifying different types of urban growth and the change dynamic in Guangzhou using multi-temporal remote sensing data
International Journal of Applied Earth Observation and Geoinformation
(2013) - et al.
The utilisation of airborne laser scanning for mapping
International Journal of Applied Earth Observation and Geoinformation
(2005)