The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
313
Fig.7 Fusion Image By PCA
Quantitative Analysis
Entropy analysis is chosen for quantitative analysis in this
study. In terns of Shannon principle, the entropy can be
calculated by equation(2)
H = -^ j p i x\og 2 p i ( 2 )
The entropies of original data and the fusion images are
calculated and listed in the following table.
image
entropy
image
entropy
Fig.l
R
7.5164
Fig.6
R
6.5786
G
7.5169
G
6.3843
B
7.4070
B
7.0040
Fig.3
5.6076
Fig.7
R
7.7382
Fig.4
4.6589
G
7.8019
Fig.5
R
6.1682
B
7.6527
G
5.9279
B
6.3698
Table.2 Entropy Contrast Table
In Table.2, the entropy values of Fig.l, Fig.2, Fig.3 shows the
information contained in the corresponding images or band
before fusion. Fig.5, Fig.6, Fig.7 represent the fusion images
by IHS transform and PCA. The entropies are listed in Table.2.
It is quite obvious that the entropy values of Fig.5 and Fig.6 is
lower than Fig.l, but higher than Fig.3 and Fig.4. It means the
spectral information in Fig.5 and Fig.6 is lost during the fusion
process from Fig.l. Of course, after fusion, the information
content have been increased from Fig.3 and Fig.4.
The entropy value of Fig.7 is higher than that of Fig.l, Fig.2
and Fig.3. It means that information in Fig.7 is more abundant
than this images. Also, the entropy values are higher than that
of Fig.5 and Fig.6. It indicates that the fusion result by PCA is
better than result by IHS transform.
3.2 Vector Fusion between LIDAR points and images
Fusion discussed in section 3.1 indicates that fusion between
raster points and Ortho-photomap. The information in these
two data sources are complemented by each other. It is also
clear that some of the spectral information is lost during the
fusion procedure. Since the data source are all in raster format,
this technology is also called raster fusion.
Here we use vector Fusion to denote the fusion process
between LIDAR points and images. When conducting the
fusion, the LIDAR points are in vector format. This technology
is used for appending the spectral information to every points
according to the location relationship. It also means that not
only coordinate data but also spectral and optical information
are included in results.
Since the coordinate relationship between points and images,
the spatial analysis tech can be used for vector fusion. The
most useful method for fusion is overlay analysis. For each
LIDAR point, the corresponding pixel which lies in the same
coordinate with the point is selected firstly. Then the spectral
data, usually in R, G and B, are acquired and attached in the
point. So, the data components of fused vector points are here
listed:
Coordinate _XCoordinate _Y,Coordinate _Z,Intensity
The forth experiment is executed for Ortho-photomap and
vector image of points cloud. Here, the I values of Ortho
photomap are replaced by the grey value of interpolation
results. Fig.5 shows the result of the first experiment.
Fig. 8 Fusion Image by Overlay Analysis Method
4. CONCLUSION
Two kinds of method is executed for fusion between airborne
laser scanning points and Ortho-photomap, raster fusion and
vector fusion. In the part of raster fusion, the IHS transform
and PCA algorithm are used for the integration of points data
which have been interpolated into raster and raster image data.
While in the part of vector fusion, the overlay analysis
technology, which is a part of spatial analysis, is used to
complete the integration of points and image. In terms of
fusion images, both of the methods is helpful. But from
quantitative analysis, the result from IHS transform is not as