| dc.description.abstract | Texture mapping for building facades is an important task in
photorealistic building modeling. Following the popularity of digital
cameras, extraction of the object texture becomes convenient. The
close-range photogrammetry, thus, can acquire rich spatial and texture
information from high overlap images. Although those close-range
images can provide detail information, occlusion is still a problem to
overcome. To derive complete textures for a block building model, this
investigation proposes a scheme using high overlap images through
geometrical and spectral analyses.
There are four parts in the proposed scheme, namely, (1)
orientation modeling, (2) generation of 3D point clouds, (3) occlusion
detection, and (4) image compensation. First step calibrates the camera
and computes orientation parameters. We acquired high overlap images
with signalized targets for camera calibration. Then, the images were
acquired for test buildings. To examine the applicability of generating
ground control points (GCPs) from building models, a small number of
structure points were extracted. In the second part, we combine CLR
(Center-Right-Left) matching and image classification to derive reliable
conjugate points for 3D point clouding. Then, geometry and spectrum are
analyzed to separate the foreground from building surfaces. On the
geometry part, space intersection is employed to calculate the object
coordinates for conjugate points. In the spectrum analyses, image
classification is performed to determine the class for each pixel.
Combining geometry and spectrum characteristics, we detect the
foreground objects for removal. The last step is image compensation. The
occluded regions in a selected master image are then replaced by the
unhidden parts extracted from slave images.
The result shows higher correctness when CLR matching and
image classification are combined. Experimental results show that the
proposed method can detect occlusion part and replaced by the unhidden
parts extracted from other images.
| en_US |