Page 27 - Surveyor 54.3 and 4

The Potential of Panoramic Images for 3D Indoor Model Reconstruction

Vol.54 No.3 & 4 2019

(a) (b) images must be at least 60%
overlapped for each pair. The
camera stations were 3-4 meters
from the interior building, this is an
ideal distance to capture the whole
surface of the since the wall is
relatively flat and it did not cause any
problems with image matching.

The photogrammetric processing
Figure 5: Panoramic images (a) Digital camera on tripod captured an image
(b) The example of images of image orientations consists of
measurement of control and tie
points and block adjustment. This
can be done with Photomodeler
based on sequential images, which enough target point to generate a software. At least, 10 - 15 tie
is quite complicated. The scope perfect 3D model. points are used to orientate one
of problems can be determined panoramic stereo pair. The standard
through the information obtained 2.3. Data Processing error of image observations has
and the effectiveness of the varied between 0.5 – 1 pixels. In
software on sequential images. The The processing method is based on PhotoModeler software, two modes
previous research helps on selecting 15 images that captured from every of the project are introduced which
the suitable area to conduct camera station. The images are are a point-based and shape-based
the research. Fakulti Senibina captured in a 180° position and in project. For this research, point-
Perancangan dan Ukur in UiTM panorama view. Figure 6 shows the based is selected. It is a process
Shah Alam has been selected as the position of camera station inside the where points are marking in the
research area for this research. The building. The purpose of capturing image to create a 3D model. Every
research area is in UiTM Shah Alam, an image in panorama view in this point that is marked on the images
Selangor Darul Ehsan. The building research is to identify the ability must be overlapping each other at
is small and rectangular. It is also of the digital camera to capture the same position on the images.
as an initial process to identify the panorama image for 3D indoor At least two images are needed to
information of object detection and modelling. This panorama scene start marking a point in the software
propose a model in 3D. is captured in sequence and it is (Figure 7). The camera information
necessary to complete 360° model.
has to be imported into the current
2.2. Data Acquisition The purpose of camera stations is research before starting any process
The images are captured in to capture and cover the panorama in PhotoModeler.
panoramic photography using view inside the building. If the The 3D modeling can be generated
digital camera Sony SLT- A35. Using images do not completely capture after the completion of data
a tripod the digital camera was set the panorama view of the building, processing. The model is described
up at the camera station. From the the orientation process will not and modelled as polylines. The
camera station, images are captured successful to be processed. The corresponding points are identified
horizontally with panoramic view. and measured in both images
The process of capturing images and the 3D coordinates are
is repeated on every five camera determined by an intersection in
stations. The whole interior building space. However, the images can
has been captured to obtain a good be viewed stereoscopically, which
result in the 3D model. ease remarkably the interpretation
of the structure. Since a panoramic
The camera captured all the interior
features of the building as shown in image sequence is rectified to a
Figure 5. The steps were repeated plane, geometrically transformed
until all 360° view are captured. If the and resampled, the resolution is
number of images is increased, then reduced. The stretching of the
the higher of overlapping features on pixels can reduce the interpretability
the images. It is important to have Figure 6: Position of camera stations as well. Therefore, it would be

27