In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010,1 APRS, Vol. XXXVIII, Part 7B 
2.1 Data registration 
First, we register the LIDAR data, aerial images and building 
models. The control points are measured and the mapping 
functions selected for registration of the three data sets to the 
same coordinate systems. There are two parts in this step, 
namely, planimetrie registration and elevation registration. The 
image coordinate system is used as reference for planimetrie 
registration, because the images provide high planimetrie 
accuracy. The LIDAR coordinate system is used as reference for 
elevation registration, because the LIDAR data provide more 
accurate elevation. Planimetrie registration note X and Y shifts 
on the plane. The elevation registration adjusts the shifts in the 
Z direction. 
Figure 1. Workflow of the propose scheme 
2.2 Change detection of 3D building models 
In this part, we examine spectrum information from aerial 
images, height difference between the LIDAR points and the 
building models and linear features of the aerial images for 
detection of different types of change. First, we use the spectral 
information from the images. Here, NDVI is calculated to detect 
the area of vegetation for the exclusion of non-building areas. 
Second, we detect the LIDAR points which represent building 
roof planes, excluding the points on the wall and the convex 
points. Here, only points within the building boundaries are 
selected to be used for Delaunay triangulation. Third, facet 
orientation analysis is carried out for each triangle to detect 
those that might include wall points. Fourth, we calculate the 
center of the circumference for the triangle. The mean value and 
standard deviation of the elevation of the points in the circle are 
then calculated. Any point in a triangle with an elevation larger 
than two standard deviations is excluded. Fifth, we use equation 
1 and the building model comer coordinates to calculate 
coplanarity parameters A, B, C. After this, the difference in 
height between the LIDAR points and the building models is 
calculated. 
Z=AX+BY+C 
(i) 
The height differences (Ah) between the LIDAR points and 
building models comprise our major information about change. 
The workflow for calculation of height difference is shown in 
Fig 2. This detection process is done model by model. Since the 
height difference is, among others, the most important factor 
considered in this study, a double-threshold strategy on that is 
proposed to cope with the high sensitivity to thresholding often 
encountered with the rule-based approach. The double 
threshold strategy distinguishes the obvious types of change 
first, so as to have more information and different thresholds to 
facilitate detection of the areas subject to further examination. 
Fig 2. Calculation of height difference 
Here, the line features in the aerial image give information that 
is used to refine the results. After this, we detect each building 
model to show the appropriate type of change. First, we set a 
double-threshold for height difference to discriminate between 
changed and unchanged points. The upper threshold is 3m and 
the low threshold is lm. The 80% (8]) points in the building 
with height difference larger than upper threshold and the 80% 
(6 2 ) points in the building with height difference smaller than 
low threshold are used to detect obvious changes and obvious 
unchanged in buildings. The data set between the double 
threshold contain the areas subject to further examination. For 
the areas subject to further examination, additional information, 
line features from aerial images are added. The workflow for 
change detection is shown in Fig 3. 
Fig 3. Change detection with double-thresholding 
Line feature comparisons facilitate detection of the areas subject 
to further examination. The process for detection on areas 
subject to further examination is shown in Fig 4. Some 
parameters are to be set. The 50% (8 3 ) points in the building 
with height difference larger than low threshold is used to detect 
the main-structure changed in buildings when the line feature 
comparisons are confirmed that there is no match. The 50% (6 4 ) 
points in the building with height difference smaller than upper 
threshold is used to detect the unchanged in buildings when the 
line feature comparisons are confirmed that there is a match.