The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008
- which have restricted their widespread adoption in the survey
and mapping industries. The system will satisfy the demand for
a mobile mapping system that can compete in both more high
position accuracy and automatic mapping, which will be
discussed in the following sections.
2. HIGH-ACCURACY POSITIONING BY FUSING
MULTI-SENSOR DATA
Due to poor navigation sensors, the absolute accuracy of the
object space points was too poor for many applications -
especially when compared with competing technologies. So for
broad utilization in more application, it must increase absolute
object space accuracies by using more high accuracy hardware
or using more sophisticated processing techniques. In this
research, the POS/LV™ of Applanix corp. is utilized as our
navigation sensor, the power of the navigation system
consumedly improve the position accuracy of mapping. But in
some of case such as in poor GPS signal area, it also cannot get
ideal accuracy.
As well known, bundle adjustment is efficient method to
improve accuracy, but it is difficult to do it full automatically
duo to fallible image matching for tie point. Based on this
knowledge, this paper presents a robust automatic extraction
method for tie points. And then, photogrammetry bundle
adjustment technology is developed to improve and stabilize
accuracy of positioning.
2.1 Robust Automatic Extraction of Tie Point
With the obtained tie points, a bundle adjustment can achieve
an improved exterior orientation of the camera, which results in
the final mapping position accuracy. To extract tie points in
image sequence robustly and precisely is the prerequisite
condition for Photogrammetry Bundle Adjustment. But,
automatic and robust extraction of tie points in image is one of
the most challenging problems in computer vision and digital
photogrammetry. The difficulty of tie point extraction is image
matching of these extracted feature points (comers), due to
easily happened amphibolous matching. The traditional and
efficient solution of the matching problem is to use rectified
images, which reduce the complexity of the matching
algorithms from 2D to ID search.
But, even if the epipolar geometry is used to image matching,
the amphibolous matching problem also happens along epipolar
line. For solving this problem perfectly, the laser 3D point
clouds are used for image matching to solve that amphibolous
matching problem. The figure 1 shows how laser 3D points
facilitate the solution of the image matching problem with any
no-amphibolous matching. With referring to 3D laser point
clouds, the extracted feature point can achieve its 3D coordinate
in mapping frame, then the initial coordinate of that feature
point in the next image can be computed by projecting 3D
coordinate to image plane. Finally, just inching adjustment is
performed by image matching. Because the adjustment is done
in small area, the amphibolous stereo image matching can be
solved and matching become more robust.
Figure 1. Automatic Tie Point Extraction by Supporting of
Laser Point Cloud
2.2 Method of Data Fusing Processing for High-Accuracy
Positioning
The key step for our high-accuracy positioning is automatic
accurately tie point extraction by supporting of laser point could,
which have been described by detail in the above section. As
the figure 2 shown, the navigation data should be calculated for
initial position and orientation for laser and camera, so that the
tie point of image can be extracted automatically with help of
the initial laser point cloud.
Figure2. Data Fusion Processing for High-accuracy
Positioning
With using accurate extracted tie points, bundle adjustment
(which is almost exclusively don in close-range
photogrammetry) is processed for improving accuracy of
camera position and orientation (X, Y, Z, <t> , a> , k ). The