Question 6: How do we merge 3D maps of the skull, face, and teeth?
The final task is the merging of data from different 3D digital maps. The conceptual basis for merging is again an idea adapted from terrestrial photogrammetry. In that domain, it is frequently desirable to merge data from overlapping 3-dimesional maps generated by aircraft or satellites surveying targets from different directions and altitudes. Figure 12 illustrates schematically the rationale for such an operation.
|
|
|
Figure 12a
|
Figure 12b
|
Detail 12a shows in plan view an aircraft taking an overlapping stereo pair of images while flying over the terrain from left to right. The region of overlap of the two images of the stereo pair, shown here shaded in red, is called the "neat" region and is analogous to the overlapping regions of Figures 7d and 7e, and also of Figures 8f and 8g. From information contained in this neat region, the 3D model shown in Detail 12b can be constructed. Note that three unambiguous landmarks -- A, B, and C -- have been located in the region of overlap in Detail 12a and have accordingly also been identified in the 3D model shown in 12b.
At a different time, a second aircraft flying over the terrain from a different direction generates another stereo pair of images (Detail 12c) from which a second 3D model can be constructed (Detail 12d).
|
|
|
Figure 12c
|
Figure 12d
|
Because the two stereo pairs were generated from different perspectives they and the 3D models generated from them do not overlap completely. However, they and the two 3D models generated from them do share in common three unambiguous landmarks: A, B, and C. If the two 3D models could be rotated and translated through space until their paired images of A, B, and C could be brought into consonance, the neat areas of the two stereo pairs could be merged to yield a continuous 3D model. A plan view of such a model is shown in Detail 12e with the neat portions of the two stereo pairs overlapping and continuous.
|
|
Figure 12e
|
|
(Please note: this is an animation. If the points arren't
blinking, check your browser's settings to make sure that
animated GIFs play continuously.)
|
At this point any landmark unambiguously identifiable on one 3D model can be precisely related to any point unambiguously identifiable on the other model. Two such points are represented by the blinking points ( and  ) in Figure 12e.
In many terrestrial mapping situations (for example, in desert areas), few or no unambiguous markers like A, B, and C common to 2 or more stereo pairs exist in the terrain. In such cases, artificial markers, called "tie points" are placed on the ground before the overflight in which the stereo images are generated.
An analogous situation exists in craniofacial imaging. Therefore we take a leaf from photogrammetrists and from the work of Bjørk, and add our own unambiguous metal markers. But in this case we fasten the tie points only temporarily to the surfaces of the teeth and face. Such an application is presented in the section on Clinical Implementation which follows.
|
Previous