Full text: Photogrammetric and remote sensing systems for data processing and analysis

  
  
Fig. 4. General stereo configuration Fig. 5. Right-to-left motion 
The epipolar lines associated with one of these epipolar planes are horizontal scanlines in the 
images (as in Figure 5). The projection of P onto these epipolar lines moves to the right as the 
camera moves to the left. The velocity of this movement along the epipolar line is a function 
of P’s distance from the line joining the lens centers. The closer the feature, the greater is its 
motion. 
For the particular motion we discuss here, the epipolar lines are not only horizontal, they occur 
at the same vertical position fscenline) in all the images. Therefore, a horizontal slice of the 
spatio-temporal data formed from this motion contains all the epipolar lines associated with 
one epipolar plane (see Figure 6). 
Figure 6 depicts three of the images sampled to form the solid of data. Typically a hundred or 
more images are used, making P's trajectory through the data a continuous path, as indicated 
in the diagram. For this type of lateral motion, if the camera moves a constant distance between 
images, the trajectories are straight lines. 
Figure 7 shows a horizontal slice through the solid of data shown in Figure 2, which was 
constructed from a sequence of 125 images taken by a camera moving from right to left. Figure 8 
shows a frontal view of that slice. We call this type of image an epipolar-plane image (EPI) 
because it is composed of one-dimensional projections of the world points lying on an epipolar 
plane. Each horizontal line of the image is one of these projections. Thus, time progresses from 
bottom to top, and, as the camera moves to the left, the features move to the right. 
       
Epipolar-Plane 
Image (EPI) 
Fig. 8. Frontal view of the EPI 
(v = 140) 
—— dá 
time 
  
Fig. 6. Sliced solid of data ^ Fig. 7. Right-to-left motion with solid 
There are several things to notice about this image. First, it contains only linear structures. 
In this respect it is much simpler than the spatial images used to create it (see Figure 1 for 
comparison). Second, the slopes of the lines determine the distances to the corresponding 
features in the world. The greater the slope, the farther the feature. Third, occlusion, which 
occurs when a closer feature moves in front of a more distant one, is immediately apparent in 
this representation. For example, the narrow white bar halfway from the left of the EPI in 
Figure 8 is initially occluded, then it is visible for a while until it is occluded briefly by a thin 
light object, then visible again before being rapidly occluded twice by two darker objects, and 
then is continuously visible until the end of the sequence. Thus, the same object is seen four 
different times. These points suggest the approach we should take in analyzing the data: find 
122 
CCE 
et.OoO OD 
m OO 
~~ 035 CSP (05
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.