|
Wavelet-Based Clustering Method to Detect Building in Urban Area from Airborne Laser Scanner Data
Vu Tuong Thuy Space Technology Applications and Research Program - Asian Institute of Technology P.O. Box 4 Klong Luang, Pathumthani, 12120, Thailand rsc009994@ait.ac.th Mitsuharu Tokunaga Associate Professor Environmental System Engineering - Kanazawa Institute of Technology 7-1,Ogigaoka, Nonoichi, Ishikawa 921-8501 Japan mtoku@neptune.kanazawa-it.ac.jp
Airborne laser scanner is an integrated system of GPS, INS and laser scanner to send the laser pulse and receive the laser hit on Earth surface. Consequently, the cloud points with X, Y and Z coordinates were acquired. After flying survey, simply speaking, the remained work to be done is to distinguish between ground surface points and object points on the surface. Nevertheless, it is not a simple processing as simply described due to the reflectance of laser hit on unpredictable distribution of objects on Earth surface. Several developed algorithms can be listed as (Haala et al. 1998) derived parameters for 3-D CAD models of basic building primitives by least-squares adjustment minimizing the distance between a laser scanning digital surface model and corresponding points on a building primitive; (Axelsson, 1999) introduced the classification algorithm based on the Minimum Description Length criterion for detection of structure; (Haala et al. 1999) integrated multi-spectral imagery and laser scanner data to extract buildings and trees in urban environment. Furthermore, one category of segmentation algorithm is the direct application on cloud points to eliminate the error introduced by interpolation such as (Maas and Vosselman, 1999) and (Sithole and Roggero, 2001). However, this type of algorithm has to pay a cost of computation time.
It is noted that most of developed algorithms is applied in European region either urban or rural sites. The distribution and the type of objects in the Asian region are completely different from the ones in European region. The obvious observation on the big Asian city is the high density of building with the interference between high buildings and very small houses. Some of developed algorithms may be successful applied in this complex scene. Originated from the mentioned point, this paper proposed the multi-resolution approach in segmentation laser points. The multi-resolution method allows the processing to analyze object at different resolutions. Therefore, it is possible to distinguish the objects of different sizes. A redundant wavelet analysis with B3 spline wavelet function (Starck and Murtagh, 1994) was utilized to form the multi-resolution method. Briefly, the wavelet-based algorithm in this paper, named wavelet-based clustering, is to smooth the cloud points, which were grid-based format, to find out the point clusters at different resolutions, and therefore, the object existences at different resolutions. Tracking the signatures over several resolutions, the required objects are easily to be detected from the rest ones. The following parts of this paper are started with the description of proposed algorithm and the testing result applied in Shinjuku area, Tokyo, Japan. Shinkjuku area is a commercial and office area, which is covered by several towels, several smaller and lower buildings, with crowded activity of people. It is a typical and interesting site to test the proposed algorithm. Conclusion and several further improved points will end up this paper. Data Processing The complete processing to detect building is summarized in the flowchart in Figure 1. 
Airborne laser scanner data used to process in this research was provided by Kokusai Kogyo Co., Ltd. Geomatics Department. This company accomplished the flying survey over most of Japan country. The parameter of the fly over our testing area is given in Table 1 below. Table 1 – Airborne laser scanner data parameters
- Interpolation: according to (Behan, 2000), the planar interpolation method on triangulated irregular network (TIN) gives the most accurate interpolated image. This interpolation method is applied here with the basic resolution of 1 meter (see Figure 2). 
Page 1 of 2
| Next | |