Wavelet-Based filtering the cloud points derived from airborne laser scanner
T. Thuy VU
Doctoral Candidate, Space Technology Applications and Research Program
Asian Institute of Technology
P.O. Box 4 Klong Luang, Pathumthani, 12120
Tel: (66)-2-524-6190
Fax: (66)-2-524-5597
Email: rsc009994@ait.ac.th
Thailand
Fumio Yamazaki
Professor, School of Civil Engineering
Asian Institute of Technology
Thailand
Mitsuharu Tokunaga
Associate Professor, Environmental System Engineering
Kanazawa Institute of Technology
Japan
Abstract
In spite of the wide application of airborne laser scanner (ALS) data in topographic mapping recently, the
segmentation of this data, usually called filtering process, is still challenging due to the complex distribution of the
objects on the earth surface. The authors propose a new method based on a wavelet-based approach to filter the
laser scanner data. Wavelet-based filtering algorithm is a detection of the clusters of the laser points across different
spatial resolutions. This enabled the classification of objects based on their size and perfectly filters out the
unwanted information at a specific resolution. The algorithm, therefore, was named ALSwave (Airborne Laser
Scanner Wavelet) method. This study is the testing process of ALSwave on a dense urban area of Tokyo, Japan. The
result showed good filtered bare earth surface coupled with the acceptable computational time. Beside a variety of
existing algorithm, wavelet-based filtering has been introduced as a potential approach to filter the cloud laser data
points and subsequently generate fast and more accurate digital terrain models.
Introduction
The major purpose of segmentation of the airborne laser scanner data can simply be described as the distinction
between the bare earth points and the overlying object points. However, objects on the earth surface are complex,
and affect the reflectance of laser hit. The existing algorithms could be categorized into two types, i.e. with or
without interpolation of the cloud point into regular grid. Filtering the grid-based laser point (Acqua, et. al., 2001;
Haala, 1999) is much faster and easier to implement but suffers from the problem of interpolation (Vosselman and
Maas, 2001). Alternatively, filtering of raw laser points could avoid the problem caused by interpolation but it pays
the cost of computation time and has hardly been implemented in recent studies (Maas and Vosselman, 1999;
Roggero, 2001; Sithole, 2001).
The algorithm proposed in this paper concentrates in another point which has not been considered in the existing
algorithms. This is the resolution (or scale) to be analyzed. Analysis of the objects in the image or the cloud points
at different resolutions has been proved as an excellent approach to detect and extract the target objects (Lega, et.
al., 1995; Starck and Murtagh, 1994). The key point is that the objects appear only at a certain range of scale, or
resolution. In this study, wavelet algorithm is introduced, for the first time, for the segmentation of airborne laser
scanner data. The idea of wavelet, originated in early 20 th century, has been attractive tool with the solid
mathematical background in the 1980’s, to several researchers and engineers of different fields. The proposed
algorithm applies the redundant a trous algorithm with B3 spline wavelet function (Starck and Murtagh, 1994), to
maintain translation invariant, which is required for feature detection. In implementation, the algorithm is a simple
convolution with 5x5 mask. The definition of this mask is presented as follow with all elements scaled up to 256.
The developed method for segmentation airborne laser data in this study is based on wavelet, and is called ALSwave
(Airborne Laser Scanner Wavelet) method. The outline of this method is presented in the following section.
Wavelet clustering algorithm for filtering airborne laser scanner data-alswave method
ALSwave is proposed as a semi-automatic algorithm to filter the airborne laser scanner data. The complete
procedure of processing is shown in flowchart (Figure 1). The step-by-step details of this algorithm are described in
the following subtopics.
Figure 1. The flowchart of ALSwave processing
