1. DSM Extraction from Pleiades Images Using RSP
Muhammad Irsyadi Firdaus, Jiann-Yeou Rau
Department of Geomatics, National Cheng Kung University, Taiwan
Introduction
The Pleiades constellation is a high-resolution satellite imagery (HRSI).
The great agility of the Pleiades satellites with their motion capabilities in a
roll, pitch, and yaw enables the system to maximize the number of
acquisitions over a certain area and provide stereos along the paths that can
obtain tri-stereo or quadri-stereo images. The purpose of this study is to
assess and measure the capacity of the Pleiades system to generate the
Digital Surface Model (DSM) using RPC Stereo Processor (RSP). RSP is a
software package for digital surface model and true-orthophoto generation
from satellite stereo imagery. In this study, DSM was produced using three
Pleiades stereo images in Taipei city. To evaluate the quality of DSM from
HRSI, we compare it with DSM generated from aerialphoto. The
comparable objects are buildings, roads, and sports fields.
Result
Conclusion
Reference
Figure 1. DSM from aerialphoto
Figure 3. DSM HRSI – DSM AP
Figure 2. DSM from Pleiades
Figure 4. True Orthophoto from
3D Model
Workflow
This paper has introduced an operational-ready software package named
RSP (RPC stereo processor) that produces DSM and true-orthophoto
from RPC-modelled stereo imagery. RSP is highly optimized and can
compute large-frame image very efficiently. In this study, the highest
DSM value difference between HRSI and Aerialphoto is on the objects
road 1, road 2 and road 3. Over the city areas, image matching was
susceptible to difficulties associated with low image contrast and
shadowing, resulting in decreased accuracy.
Qin, Rongjun. 2016. RPC Stereo Processor (RSP) – A Software Package
For Digital Surface Model And Orthophoto Generation From
Satellite Stereo Imagery. ISPRS Annals of the Photogrammetry,
Remote Sensing and Spatial Information Sciences, Volume III-1.
Qin, R., A. Gruen and C. Fraser, 2014. Quality Assessment of Image
Matchers for DSM Generation - A Comparative Study Based on
UAV Images. In: Asian Conference on Remote Sensing, Nay Pyi
Taw, Myanmar, 27-31, Octomber, 2014, pp.
DSM AP Building 1 DSM HRSI Building 1
DSM AP Road 2 DSM HRSI Road 2
Figure 9. Graphic Difference DSM HRSI - DSM AP
In order to take into account the influence of topographic variation on the
modelled surface, the DSM strip area was divided into separate sub-areas
representing different land covers. These comprised urban areas, with a
further subdivision into regions within these three categories. DSM HRSI-AP
Building 1
DSM HRSI-AP
Road 2
-20
-15
-10
-5
0
5
10
15
20
Building 1 Building 2 Building 3 Building 4 Building 5 Road 1 Road 2 Road 3 Sports 1 Sports 2
Difference DSM HRSI – DSM AP
Difference DSM HRSI - DSM AP Min Difference DSM HRSI - DSM AP Max
Difference DSM HRSI - DSM AP Mean Difference DSM HRSI - DSM AP StdDev
In Fig. 9, the urban area has three categories incorporated the building class
including tall buildings; sporting fields (Sporting fields class)
encompassing flat tennis courts and basketball court; and roadway (road
class) encompassing crossroads.
The results are summarized in Fig. 9 and show that, in general, accuracy
degrades in road class. Categories that have low accuracy differences
are buildings and sports.
Fig. 6. Building 1 APFig. 5. Building 1 HRSI
Fig. 8. Road 1 APFig. 7. Road 1 HRSI
brightness and contrast while in figure 6 is building obtained from
aerialphoto which has higher brightness and contrast.
The image difference
between figure 5 and
6 is in brightness and
contrast images.
Figure 5 is building
1 obtained from
HRSI which has low
shadows when compared with image 8 which has no shadow and a higher
contrast image. This will affect the image matching and accuracy of DSM.
For the road
category, in figure 7
and figure 8 is a
road 1 image that
has a very clear
distinction. Figure 7
has low contrast,
RPC Stereo Pair
DSM and True-Orthophoto
Image Correction
Bundle adjustment
Dense Image Matching
DSM and True-Orthophoto
DSM Overlay
Analysis
RSP
Tie-point Matching
Bundle adjustment
Dense Image Matching
Aerial Images