CMRT09: Object Extraction for 3D City Models, Road Databases and Traffic Monitoring - Concepts, Algorithms, and Evaluation 
e f 
Figure 2. a) Manually digitized objects in residential area, b) 
Manually digitized objects in industrial area, c) Clustering 
results of /r-means in residential area, d) Clustering results of 
Ar-means in industrial area, e) Clustering results of artificial 
swarm bee colony algorithm in residential area, f) Clustering 
results of swarm bee algorithm in industrial area. 
• Local height variation which is computed using a small 
window (3*3) around a data sample. 
• Last echo intensity 
Figure 1. a) Aerial image of residential area, b) Aerial image of 
industrial area, c) First echo LIDAR range data of residential 
area, d) First echo LIDAR range data of industrial area, e) Last 
echo LIDAR range data of residential area, f) Last echo LIDAR 
range data of industrial area, g) Overlaid of manually digitized 
objects in residential area; h) Overlaid of manually digitized 
objects in residential area 
The normalized difference of the first and last echo range 
images is used as the major feature band for discrimination of 
the vegetation pixels from the others. According to the 
above defined features, the &-means and artificial 
swarm bee algorithm were developed based on the 
parameters listed in table 1. 
The first step in every clustering process is to extract the feature 
image bands. The features of theses feature bands should carry 
useful textural or surface related information to differentiate 
between regions related to the surface. Several features have 
been proposed for clustering of range data. Axelsson (1999) 
employs the second derivatives to find textural variations and 
Maas (1999) utilizes a feature vector including the original 
height data, the Laplace operator, maximum slope measures and 
others in order to classify the data. In the following 
experiments we used five types of features: 
• LIDAR range data 
• The difference between first and last echo range images 
• Top-Hat filtered last echo range image 
Table 1. Parameters used in the clustering of LIDAR datasets 
Algorithm 
Parameters Value 
Ar-means 
Maximum number of iterations 
1000 
Number of scout bees, n 
35 
Number of sites selected for neighbourhood 
11 
search,m 
Artificial 
Number of best “elite” sites out of m 
swarm bee 
selected sites, e 
1 
colony 
Number of bees recruited for best e sites, 
algorithm 
nep 
/ 
Number of bees recruited for the other (m- 
e) selected sites, nsp 
J 
Number of iterations, R 
200